git/parallel-checkout.c

673 lines
18 KiB
C
Raw Normal View History

unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
#include "cache.h"
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
#include "config.h"
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
#include "entry.h"
#include "parallel-checkout.h"
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
#include "pkt-line.h"
#include "progress.h"
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
#include "run-command.h"
#include "sigchain.h"
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
#include "streaming.h"
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
#include "thread-utils.h"
#include "trace2.h"
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
struct pc_worker {
struct child_process cp;
size_t next_item_to_complete, nr_items_to_complete;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
};
struct parallel_checkout {
enum pc_status status;
struct parallel_checkout_item *items; /* The parallel checkout queue. */
size_t nr, alloc;
struct progress *progress;
unsigned int *progress_cnt;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
};
static struct parallel_checkout parallel_checkout;
enum pc_status parallel_checkout_status(void)
{
return parallel_checkout.status;
}
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
static const int DEFAULT_THRESHOLD_FOR_PARALLELISM = 100;
static const int DEFAULT_NUM_WORKERS = 1;
void get_parallel_checkout_configs(int *num_workers, int *threshold)
{
char *env_workers = getenv("GIT_TEST_CHECKOUT_WORKERS");
if (env_workers && *env_workers) {
if (strtol_i(env_workers, 10, num_workers)) {
die("invalid value for GIT_TEST_CHECKOUT_WORKERS: '%s'",
env_workers);
}
if (*num_workers < 1)
*num_workers = online_cpus();
*threshold = 0;
return;
}
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
if (git_config_get_int("checkout.workers", num_workers))
*num_workers = DEFAULT_NUM_WORKERS;
else if (*num_workers < 1)
*num_workers = online_cpus();
if (git_config_get_int("checkout.thresholdForParallelism", threshold))
*threshold = DEFAULT_THRESHOLD_FOR_PARALLELISM;
}
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
void init_parallel_checkout(void)
{
if (parallel_checkout.status != PC_UNINITIALIZED)
BUG("parallel checkout already initialized");
parallel_checkout.status = PC_ACCEPTING_ENTRIES;
}
static void finish_parallel_checkout(void)
{
if (parallel_checkout.status == PC_UNINITIALIZED)
BUG("cannot finish parallel checkout: not initialized yet");
free(parallel_checkout.items);
memset(&parallel_checkout, 0, sizeof(parallel_checkout));
}
static int is_eligible_for_parallel_checkout(const struct cache_entry *ce,
const struct conv_attrs *ca)
{
enum conv_attrs_classification c;
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
size_t packed_item_size;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
/*
* Symlinks cannot be checked out in parallel as, in case of path
* collision, they could racily replace leading directories of other
* entries being checked out. Submodules are checked out in child
* processes, which have their own parallel checkout queues.
*/
if (!S_ISREG(ce->ce_mode))
return 0;
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
packed_item_size = sizeof(struct pc_item_fixed_portion) + ce->ce_namelen +
(ca->working_tree_encoding ? strlen(ca->working_tree_encoding) : 0);
/*
* The amount of data we send to the workers per checkout item is
* typically small (75~300B). So unless we find an insanely huge path
* of 64KB, we should never reach the 65KB limit of one pkt-line. If
* that does happen, we let the sequential code handle the item.
*/
if (packed_item_size > LARGE_PACKET_DATA_MAX)
return 0;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
c = classify_conv_attrs(ca);
switch (c) {
case CA_CLASS_INCORE:
return 1;
case CA_CLASS_INCORE_FILTER:
/*
* It would be safe to allow concurrent instances of
* single-file smudge filters, like rot13, but we should not
* assume that all filters are parallel-process safe. So we
* don't allow this.
*/
return 0;
case CA_CLASS_INCORE_PROCESS:
/*
* The parallel queue and the delayed queue are not compatible,
* so they must be kept completely separated. And we can't tell
* if a long-running process will delay its response without
* actually asking it to perform the filtering. Therefore, this
* type of filter is not allowed in parallel checkout.
*
* Furthermore, there should only be one instance of the
* long-running process filter as we don't know how it is
* managing its own concurrency. So, spreading the entries that
* requisite such a filter among the parallel workers would
* require a lot more inter-process communication. We would
* probably have to designate a single process to interact with
* the filter and send all the necessary data to it, for each
* entry.
*/
return 0;
case CA_CLASS_STREAMABLE:
return 1;
default:
BUG("unsupported conv_attrs classification '%d'", c);
}
}
int enqueue_checkout(struct cache_entry *ce, struct conv_attrs *ca)
{
struct parallel_checkout_item *pc_item;
if (parallel_checkout.status != PC_ACCEPTING_ENTRIES ||
!is_eligible_for_parallel_checkout(ce, ca))
return -1;
ALLOC_GROW(parallel_checkout.items, parallel_checkout.nr + 1,
parallel_checkout.alloc);
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
pc_item = &parallel_checkout.items[parallel_checkout.nr];
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
pc_item->ce = ce;
memcpy(&pc_item->ca, ca, sizeof(pc_item->ca));
pc_item->status = PC_ITEM_PENDING;
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
pc_item->id = parallel_checkout.nr;
parallel_checkout.nr++;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
return 0;
}
size_t pc_queue_size(void)
{
return parallel_checkout.nr;
}
static void advance_progress_meter(void)
{
if (parallel_checkout.progress) {
(*parallel_checkout.progress_cnt)++;
display_progress(parallel_checkout.progress,
*parallel_checkout.progress_cnt);
}
}
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
static int handle_results(struct checkout *state)
{
int ret = 0;
size_t i;
int have_pending = 0;
/*
* We first update the successfully written entries with the collected
* stat() data, so that they can be found by mark_colliding_entries(),
* in the next loop, when necessary.
*/
for (i = 0; i < parallel_checkout.nr; i++) {
struct parallel_checkout_item *pc_item = &parallel_checkout.items[i];
if (pc_item->status == PC_ITEM_WRITTEN)
update_ce_after_write(state, pc_item->ce, &pc_item->st);
}
for (i = 0; i < parallel_checkout.nr; i++) {
struct parallel_checkout_item *pc_item = &parallel_checkout.items[i];
switch(pc_item->status) {
case PC_ITEM_WRITTEN:
/* Already handled */
break;
case PC_ITEM_COLLIDED:
/*
* The entry could not be checked out due to a path
* collision with another entry. Since there can only
* be one entry of each colliding group on the disk, we
* could skip trying to check out this one and move on.
* However, this would leave the unwritten entries with
* null stat() fields on the index, which could
* potentially slow down subsequent operations that
* require refreshing it: git would not be able to
* trust st_size and would have to go to the filesystem
* to see if the contents match (see ie_modified()).
*
* Instead, let's pay the overhead only once, now, and
* call checkout_entry_ca() again for this file, to
* have its stat() data stored in the index. This also
* has the benefit of adding this entry and its
* colliding pair to the collision report message.
* Additionally, this overwriting behavior is consistent
* with what the sequential checkout does, so it doesn't
* add any extra overhead.
*/
ret |= checkout_entry_ca(pc_item->ce, &pc_item->ca,
state, NULL, NULL);
advance_progress_meter();
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
break;
case PC_ITEM_PENDING:
have_pending = 1;
/* fall through */
case PC_ITEM_FAILED:
ret = -1;
break;
default:
BUG("unknown checkout item status in parallel checkout");
}
}
if (have_pending)
error("parallel checkout finished with pending entries");
return ret;
}
static int reset_fd(int fd, const char *path)
{
if (lseek(fd, 0, SEEK_SET) != 0)
return error_errno("failed to rewind descriptor of '%s'", path);
if (ftruncate(fd, 0))
return error_errno("failed to truncate file '%s'", path);
return 0;
}
static int write_pc_item_to_fd(struct parallel_checkout_item *pc_item, int fd,
const char *path)
{
int ret;
struct stream_filter *filter;
struct strbuf buf = STRBUF_INIT;
char *blob;
unsigned long size;
ssize_t wrote;
/* Sanity check */
assert(is_eligible_for_parallel_checkout(pc_item->ce, &pc_item->ca));
filter = get_stream_filter_ca(&pc_item->ca, &pc_item->ce->oid);
if (filter) {
if (stream_blob_to_fd(fd, &pc_item->ce->oid, filter, 1)) {
/* On error, reset fd to try writing without streaming */
if (reset_fd(fd, path))
return -1;
} else {
return 0;
}
}
blob = read_blob_entry(pc_item->ce, &size);
if (!blob)
return error("cannot read object %s '%s'",
oid_to_hex(&pc_item->ce->oid), pc_item->ce->name);
/*
* checkout metadata is used to give context for external process
* filters. Files requiring such filters are not eligible for parallel
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
* checkout, so pass NULL. Note: if that changes, the metadata must also
* be passed from the main process to the workers.
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
*/
ret = convert_to_working_tree_ca(&pc_item->ca, pc_item->ce->name,
blob, size, &buf, NULL);
if (ret) {
size_t newsize;
free(blob);
blob = strbuf_detach(&buf, &newsize);
size = newsize;
}
wrote = write_in_full(fd, blob, size);
free(blob);
if (wrote < 0)
return error("unable to write file '%s'", path);
return 0;
}
static int close_and_clear(int *fd)
{
int ret = 0;
if (*fd >= 0) {
ret = close(*fd);
*fd = -1;
}
return ret;
}
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
void write_pc_item(struct parallel_checkout_item *pc_item,
struct checkout *state)
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
{
unsigned int mode = (pc_item->ce->ce_mode & 0100) ? 0777 : 0666;
int fd = -1, fstat_done = 0;
struct strbuf path = STRBUF_INIT;
const char *dir_sep;
strbuf_add(&path, state->base_dir, state->base_dir_len);
strbuf_add(&path, pc_item->ce->name, pc_item->ce->ce_namelen);
dir_sep = find_last_dir_sep(path.buf);
/*
* The leading dirs should have been already created by now. But, in
* case of path collisions, one of the dirs could have been replaced by
* a symlink (checked out after we enqueued this entry for parallel
* checkout). Thus, we must check the leading dirs again.
*/
if (dir_sep && !has_dirs_only_path(path.buf, dir_sep - path.buf,
state->base_dir_len)) {
pc_item->status = PC_ITEM_COLLIDED;
trace2_data_string("pcheckout", NULL, "collision/dirname", path.buf);
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
goto out;
}
fd = open(path.buf, O_WRONLY | O_CREAT | O_EXCL, mode);
if (fd < 0) {
if (errno == EEXIST || errno == EISDIR) {
/*
* Errors which probably represent a path collision.
* Suppress the error message and mark the item to be
* retried later, sequentially. ENOTDIR and ENOENT are
* also interesting, but the above has_dirs_only_path()
* call should have already caught these cases.
*/
pc_item->status = PC_ITEM_COLLIDED;
trace2_data_string("pcheckout", NULL,
"collision/basename", path.buf);
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
} else {
error_errno("failed to open file '%s'", path.buf);
pc_item->status = PC_ITEM_FAILED;
}
goto out;
}
if (write_pc_item_to_fd(pc_item, fd, path.buf)) {
/* Error was already reported. */
pc_item->status = PC_ITEM_FAILED;
close_and_clear(&fd);
unlink(path.buf);
goto out;
}
fstat_done = fstat_checkout_output(fd, state, &pc_item->st);
if (close_and_clear(&fd)) {
error_errno("unable to close file '%s'", path.buf);
pc_item->status = PC_ITEM_FAILED;
goto out;
}
if (state->refresh_cache && !fstat_done && lstat(path.buf, &pc_item->st) < 0) {
error_errno("unable to stat just-written file '%s'", path.buf);
pc_item->status = PC_ITEM_FAILED;
goto out;
}
pc_item->status = PC_ITEM_WRITTEN;
out:
strbuf_release(&path);
}
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
static void send_one_item(int fd, struct parallel_checkout_item *pc_item)
{
size_t len_data;
char *data, *variant;
struct pc_item_fixed_portion *fixed_portion;
const char *working_tree_encoding = pc_item->ca.working_tree_encoding;
size_t name_len = pc_item->ce->ce_namelen;
size_t working_tree_encoding_len = working_tree_encoding ?
strlen(working_tree_encoding) : 0;
/*
* Any changes in the calculation of the message size must also be made
* in is_eligible_for_parallel_checkout().
*/
len_data = sizeof(struct pc_item_fixed_portion) + name_len +
working_tree_encoding_len;
data = xmalloc(len_data);
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
fixed_portion = (struct pc_item_fixed_portion *)data;
fixed_portion->id = pc_item->id;
fixed_portion->ce_mode = pc_item->ce->ce_mode;
fixed_portion->crlf_action = pc_item->ca.crlf_action;
fixed_portion->ident = pc_item->ca.ident;
fixed_portion->name_len = name_len;
fixed_portion->working_tree_encoding_len = working_tree_encoding_len;
/*
* We pad the unused bytes in the hash array because, otherwise,
* Valgrind would complain about passing uninitialized bytes to a
* write() syscall. The warning doesn't represent any real risk here,
* but it could hinder the detection of actual errors.
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
*/
oidcpy_with_padding(&fixed_portion->oid, &pc_item->ce->oid);
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
variant = data + sizeof(*fixed_portion);
if (working_tree_encoding_len) {
memcpy(variant, working_tree_encoding, working_tree_encoding_len);
variant += working_tree_encoding_len;
}
memcpy(variant, pc_item->ce->name, name_len);
packet_write(fd, data, len_data);
free(data);
}
static void send_batch(int fd, size_t start, size_t nr)
{
size_t i;
sigchain_push(SIGPIPE, SIG_IGN);
for (i = 0; i < nr; i++)
send_one_item(fd, &parallel_checkout.items[start + i]);
packet_flush(fd);
sigchain_pop(SIGPIPE);
}
static struct pc_worker *setup_workers(struct checkout *state, int num_workers)
{
struct pc_worker *workers;
int i, workers_with_one_extra_item;
size_t base_batch_size, batch_beginning = 0;
ALLOC_ARRAY(workers, num_workers);
for (i = 0; i < num_workers; i++) {
struct child_process *cp = &workers[i].cp;
child_process_init(cp);
cp->git_cmd = 1;
cp->in = -1;
cp->out = -1;
cp->clean_on_exit = 1;
strvec_push(&cp->args, "checkout--worker");
if (state->base_dir_len)
strvec_pushf(&cp->args, "--prefix=%s", state->base_dir);
if (start_command(cp))
die("failed to spawn checkout worker");
}
base_batch_size = parallel_checkout.nr / num_workers;
workers_with_one_extra_item = parallel_checkout.nr % num_workers;
for (i = 0; i < num_workers; i++) {
struct pc_worker *worker = &workers[i];
size_t batch_size = base_batch_size;
/* distribute the extra work evenly */
if (i < workers_with_one_extra_item)
batch_size++;
send_batch(worker->cp.in, batch_beginning, batch_size);
worker->next_item_to_complete = batch_beginning;
worker->nr_items_to_complete = batch_size;
batch_beginning += batch_size;
}
return workers;
}
static void finish_workers(struct pc_worker *workers, int num_workers)
{
int i;
/*
* Close pipes before calling finish_command() to let the workers
* exit asynchronously and avoid spending extra time on wait().
*/
for (i = 0; i < num_workers; i++) {
struct child_process *cp = &workers[i].cp;
if (cp->in >= 0)
close(cp->in);
if (cp->out >= 0)
close(cp->out);
}
for (i = 0; i < num_workers; i++) {
int rc = finish_command(&workers[i].cp);
if (rc > 128) {
/*
* For a normal non-zero exit, the worker should have
* already printed something useful to stderr. But a
* death by signal should be mentioned to the user.
*/
error("checkout worker %d died of signal %d", i, rc - 128);
}
}
free(workers);
}
static inline void assert_pc_item_result_size(int got, int exp)
{
if (got != exp)
BUG("wrong result size from checkout worker (got %dB, exp %dB)",
got, exp);
}
static void parse_and_save_result(const char *buffer, int len,
struct pc_worker *worker)
{
struct pc_item_result *res;
struct parallel_checkout_item *pc_item;
struct stat *st = NULL;
if (len < PC_ITEM_RESULT_BASE_SIZE)
BUG("too short result from checkout worker (got %dB, exp >=%dB)",
len, (int)PC_ITEM_RESULT_BASE_SIZE);
res = (struct pc_item_result *)buffer;
/*
* Worker should send either the full result struct on success, or
* just the base (i.e. no stat data), otherwise.
*/
if (res->status == PC_ITEM_WRITTEN) {
assert_pc_item_result_size(len, (int)sizeof(struct pc_item_result));
st = &res->st;
} else {
assert_pc_item_result_size(len, (int)PC_ITEM_RESULT_BASE_SIZE);
}
if (!worker->nr_items_to_complete)
BUG("received result from supposedly finished checkout worker");
if (res->id != worker->next_item_to_complete)
BUG("unexpected item id from checkout worker (got %"PRIuMAX", exp %"PRIuMAX")",
(uintmax_t)res->id, (uintmax_t)worker->next_item_to_complete);
worker->next_item_to_complete++;
worker->nr_items_to_complete--;
pc_item = &parallel_checkout.items[res->id];
pc_item->status = res->status;
if (st)
pc_item->st = *st;
if (res->status != PC_ITEM_COLLIDED)
advance_progress_meter();
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
}
static void gather_results_from_workers(struct pc_worker *workers,
int num_workers)
{
int i, active_workers = num_workers;
struct pollfd *pfds;
CALLOC_ARRAY(pfds, num_workers);
for (i = 0; i < num_workers; i++) {
pfds[i].fd = workers[i].cp.out;
pfds[i].events = POLLIN;
}
while (active_workers) {
int nr = poll(pfds, num_workers, -1);
if (nr < 0) {
if (errno == EINTR)
continue;
die_errno("failed to poll checkout workers");
}
for (i = 0; i < num_workers && nr > 0; i++) {
struct pc_worker *worker = &workers[i];
struct pollfd *pfd = &pfds[i];
if (!pfd->revents)
continue;
if (pfd->revents & POLLIN) {
int len = packet_read(pfd->fd, NULL, NULL,
packet_buffer,
sizeof(packet_buffer), 0);
if (len < 0) {
BUG("packet_read() returned negative value");
} else if (!len) {
pfd->fd = -1;
active_workers--;
} else {
parse_and_save_result(packet_buffer,
len, worker);
}
} else if (pfd->revents & POLLHUP) {
pfd->fd = -1;
active_workers--;
} else if (pfd->revents & (POLLNVAL | POLLERR)) {
die("error polling from checkout worker");
}
nr--;
}
}
free(pfds);
}
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
static void write_items_sequentially(struct checkout *state)
{
size_t i;
for (i = 0; i < parallel_checkout.nr; i++) {
struct parallel_checkout_item *pc_item = &parallel_checkout.items[i];
write_pc_item(pc_item, state);
if (pc_item->status != PC_ITEM_COLLIDED)
advance_progress_meter();
}
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
}
int run_parallel_checkout(struct checkout *state, int num_workers, int threshold,
struct progress *progress, unsigned int *progress_cnt)
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
{
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
int ret;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
if (parallel_checkout.status != PC_ACCEPTING_ENTRIES)
BUG("cannot run parallel checkout: uninitialized or already running");
parallel_checkout.status = PC_RUNNING;
parallel_checkout.progress = progress;
parallel_checkout.progress_cnt = progress_cnt;
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
if (parallel_checkout.nr < num_workers)
num_workers = parallel_checkout.nr;
parallel-checkout: add configuration options Make parallel checkout configurable by introducing two new settings: checkout.workers and checkout.thresholdForParallelism. The first defines the number of workers (where one means sequential checkout), and the second defines the minimum number of entries to attempt parallel checkout. To decide the default value for checkout.workers, the parallel version was benchmarked during three operations in the linux repo, with cold cache: cloning v5.8, checking out v5.8 from v2.6.15 (checkout I) and checking out v5.8 from v5.7 (checkout II). The four tables below show the mean run times and standard deviations for 5 runs in: a local file system on SSD, a local file system on HDD, a Linux NFS server, and Amazon EFS (all on Linux). Each parallel checkout test was executed with the number of workers that brings the best overall results in that environment. Local SSD: Sequential 10 workers Speedup Clone 8.805 s ± 0.043 s 3.564 s ± 0.041 s 2.47 ± 0.03 Checkout I 9.678 s ± 0.057 s 4.486 s ± 0.050 s 2.16 ± 0.03 Checkout II 5.034 s ± 0.072 s 3.021 s ± 0.038 s 1.67 ± 0.03 Local HDD: Sequential 10 workers Speedup Clone 32.288 s ± 0.580 s 30.724 s ± 0.522 s 1.05 ± 0.03 Checkout I 54.172 s ± 7.119 s 54.429 s ± 6.738 s 1.00 ± 0.18 Checkout II 40.465 s ± 2.402 s 38.682 s ± 1.365 s 1.05 ± 0.07 Linux NFS server (v4.1, on EBS, single availability zone): Sequential 32 workers Speedup Clone 240.368 s ± 6.347 s 57.349 s ± 0.870 s 4.19 ± 0.13 Checkout I 242.862 s ± 2.215 s 58.700 s ± 0.904 s 4.14 ± 0.07 Checkout II 65.751 s ± 1.577 s 23.820 s ± 0.407 s 2.76 ± 0.08 EFS (v4.1, replicated over multiple availability zones): Sequential 32 workers Speedup Clone 922.321 s ± 2.274 s 210.453 s ± 3.412 s 4.38 ± 0.07 Checkout I 1011.300 s ± 7.346 s 297.828 s ± 0.964 s 3.40 ± 0.03 Checkout II 294.104 s ± 1.836 s 126.017 s ± 1.190 s 2.33 ± 0.03 The above benchmarks show that parallel checkout is most effective on repositories located on an SSD or over a distributed file system. For local file systems on spinning disks, and/or older machines, the parallelism does not always bring a good performance. For this reason, the default value for checkout.workers is one, a.k.a. sequential checkout. To decide the default value for checkout.thresholdForParallelism, another benchmark was executed in the "Local SSD" setup, where parallel checkout showed to be beneficial. This time, we compared the runtime of a `git checkout -f`, with and without parallelism, after randomly removing an increasing number of files from the Linux working tree. The "sequential fallback" column below corresponds to the executions where checkout.workers was 10 but checkout.thresholdForParallelism was equal to the number of to-be-updated files plus one (so that we end up writing sequentially). Each test case was sampled 15 times, and each sample had a randomly different set of files removed. Here are the results: sequential fallback 10 workers speedup 10 files 772.3 ms ± 12.6 ms 769.0 ms ± 13.6 ms 1.00 ± 0.02 20 files 780.5 ms ± 15.8 ms 775.2 ms ± 9.2 ms 1.01 ± 0.02 50 files 806.2 ms ± 13.8 ms 767.4 ms ± 8.5 ms 1.05 ± 0.02 100 files 833.7 ms ± 21.4 ms 750.5 ms ± 16.8 ms 1.11 ± 0.04 200 files 897.6 ms ± 30.9 ms 730.5 ms ± 14.7 ms 1.23 ± 0.05 500 files 1035.4 ms ± 48.0 ms 677.1 ms ± 22.3 ms 1.53 ± 0.09 1000 files 1244.6 ms ± 35.6 ms 654.0 ms ± 38.3 ms 1.90 ± 0.12 2000 files 1488.8 ms ± 53.4 ms 658.8 ms ± 23.8 ms 2.26 ± 0.12 From the above numbers, 100 files seems to be a reasonable default value for the threshold setting. Note: Up to 1000 files, we observe a drop in the execution time of the parallel code with an increase in the number of files. This is a rather odd behavior, but it was observed in multiple repetitions. Above 1000 files, the execution time increases according to the number of files, as one would expect. About the test environments: Local SSD tests were executed on an i7-7700HQ (4 cores with hyper-threading) running Manjaro Linux. Local HDD tests were executed on an Intel(R) Xeon(R) E3-1230 (also 4 cores with hyper-threading), HDD Seagate Barracuda 7200.14 SATA 3.1, running Debian. NFS and EFS tests were executed on an Amazon EC2 c5n.xlarge instance, with 4 vCPUs. The Linux NFS server was running on a m6g.large instance with 2 vCPUSs and a 1 TB EBS GP2 volume. Before each timing, the linux repository was removed (or checked out back to its previous state), and `sync && sysctl vm.drop_caches=3` was executed. Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:55 +00:00
if (num_workers <= 1 || parallel_checkout.nr < threshold) {
parallel-checkout: make it truly parallel Use multiple worker processes to distribute the queued entries and call write_pc_item() in parallel for them. The items are distributed uniformly in contiguous chunks. This minimizes the chances of two workers writing to the same directory simultaneously, which could affect performance due to lock contention in the kernel. Work stealing (or any other format of re-distribution) is not implemented yet. The protocol between the main process and the workers is quite simple. They exchange binary messages packed in pkt-line format, and use PKT-FLUSH to mark the end of input (from both sides). The main process starts the communication by sending N pkt-lines, each corresponding to an item that needs to be written. These packets contain all the necessary information to load, smudge, and write the blob associated with each item. Then it waits for the worker to send back N pkt-lines containing the results for each item. The resulting packet must contain: the identification number of the item that it refers to, the status of the operation, and the lstat() data gathered after writing the file (iff the operation was successful). For now, checkout always uses a hardcoded value of 2 workers, only to demonstrate that the parallel checkout framework correctly divides and writes the queued entries. The next patch will add user configurations and define a more reasonable default, based on tests with the said settings. Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:54 +00:00
write_items_sequentially(state);
} else {
struct pc_worker *workers = setup_workers(state, num_workers);
gather_results_from_workers(workers, num_workers);
finish_workers(workers, num_workers);
}
unpack-trees: add basic support for parallel checkout This new interface allows us to enqueue some of the entries being checked out to later uncompress them, apply in-process filters, and write out the files in parallel. For now, the parallel checkout machinery is enabled by default and there is no user configuration, but run_parallel_checkout() just writes the queued entries in sequence (without spawning additional workers). The next patch will actually implement the parallelism and, later, we will make it configurable. Note that, to avoid potential data races, not all entries are eligible for parallel checkout. Also, paths that collide on disk (e.g. case-sensitive paths in case-insensitive file systems), are detected by the parallel checkout code and skipped, so that they can be safely sequentially handled later. The collision detection works like the following: - If the collision was at basename (e.g. 'a/b' and 'a/B'), the framework detects it by looking for EEXIST and EISDIR errors after an open(O_CREAT | O_EXCL) failure. - If the collision was at dirname (e.g. 'a/b' and 'A'), it is detected at the has_dirs_only_path() check, which is done for the leading path of each item in the parallel checkout queue. Both verifications rely on the fact that, before enqueueing an entry for parallel checkout, checkout_entry() makes sure that there is no file at the entry's path and that its leading components are all real directories. So, any later change in these conditions indicates that there was a collision (either between two parallel-eligible entries or between an eligible and an ineligible one). After all parallel-eligible entries have been processed, the collided (and thus, skipped) entries are sequentially fed to checkout_entry() again. This is similar to the way the current code deals with collisions, overwriting the previously checked out entries with the subsequent ones. The only difference is that, since we no longer create the files in the same order that they appear on index, we are not able to determine which of the colliding entries will survive on disk (for the classic code, it is always the last entry). Co-authored-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Co-authored-by: Jeff Hostetler <jeffhost@microsoft.com> Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-19 00:14:53 +00:00
ret = handle_results(state);
finish_parallel_checkout();
return ret;
}