Sorry, you need to enable JavaScript to visit this website.

Linux Kernel Performance

Linux development evolves rapidly. The performance and scalability of the OS kernel has been a key part of its success. However, discussions have appeared on LKML (Linux Kernel Mailing List) regarding large performance regression between kernel versions. These discussions underscore the need for a systematic and disciplined way to characterize, improve, and test Linux kernel performance. Our goal is to work with the Linux community to further enhance the Linux kernel with consistent performance increases (avoiding degradations) across releases. The information available on this site gives community members better information about what 0-Day and LKP (Linux Kernel Performance) are doing to preserve performance integrity of the kernel.

0-DAY CI LINUX KERNEL PERFORMANCE REPORT (V5.11)

BY Beibei Si ON Mar 02, 2021

0-Day CI Linux Kernel Performance Report (v5.11)

Contents

1.Introduction

0-Day CI is an automated Linux kernel test service that provides comprehensive test coverage of the Linux kernel. It covers kernel build, static analysis, boot, functional, performance and power tests. This report shows the recent observations of kernel performance status on IA platform based on the test results from 0-Day CI service. It is structured in the following manner:

  • Section 2, test parameter description
  • Section 3, merged regressions and improvements in v5.11 release candidates
  • Section 4, captured regressions and improvements by shift-left testing during developers’ and maintainers’ tree during v5.11 release cycle
  • Section 5, performance comparison among different kernel releases
  • Section 6, test machine list

 

2.Test parameters descriptions

Here are the descriptions for each parameter/field used in the tests.

 

Classification

Name

Description

General

runtime

Run the test case within a certain time period (seconds or minutes)

 

nr_task

If it is an integer, which means the number of processes/threads (to run the workload) of this job. Default is 1.

If it is a percentage, e.g. 200% means the number of processes/threads is double of cpu number

 

nr_threads

Alias of nr_task

 

iterations

Number to repeat this job

 

test_size

Test disk size or memory size

 

set_nic_irq_affinity

Set NIC interrupt affinity

 

disable_latency_stats

Latency_stats may introduce too much noise if there are too many context switches, allow to disable it

 

 

transparent_hugepage

Set transparent hugepage policy (/sys/kernel/mm/transparent_hugepage)

 

boot_params:bp1_memmap

Boot parameters of memmap

 

disk:nr_pmem

number of pmem partitions used by test

 

swap:priority

Priority means the  priority  of  the  swap device. priority is a value between -1 and 32767, the default is -1 and higher priority with higher value.

Test Machine

model

Name of Intel processor microarchitecture

 

brand

Brand name of cpu

 

cpu_number

Number of cpu

 

memory

Size of memory

 

 

3.Linux Kernel v5.11 Release Test

The 5.11 release of the Linux kernel was on February 14, 2021. Linus has released the 5.11 kernel. "I know it's Valentine's Day here in the US - maybe give this release a good testing before you go back and play with development kernels. All right? Because I'm sure your SO will understand." Headline features in 5.11 include Intel SGX support, a new system-call interception mechanism, the seccomp() constant-action bitmap optimization, the internal kmap_local() API, the epoll_pwait2() system call, and much more. See the LWN merge-window articles (part 1part 2) and the (under development) KernelNewbies 5.11 page for more information.

 

0-Day CI monitored the release closely to trace down the performance status on IA platform. 0-Day observed 7 regressions and 8 improvements during feature development phase for v5.11. We will share more detailed information together with correlated patches that led to the results. Note that the assessment is limited by the test coverage 0-Day has now. The list is summarized in the observation summary section.

3.1.Observation Summary

0-Day CI observed 7 regressions and 8 improvements during the feature development phase for v5.11, which is in the time frame from v5.11-rc1 to v5.11 release.

Test Indicator

Report

Test Scenario

Test Machine

Development Base

Status

fio.write_iops

[btrfs] e076ab2a2c: -18.3% regression

1SSD-btrfs-300s-8-randwrite-4k-sync-256g-performance- disk: 1SSD

fs: btrfs

runtime: 300s

nr_task: 8

rw: randwrite

bs: 4k

ioengine: sync

test_size: 256g

cpufreq_governor: performance

lkp-csl-2ap1

v5.10-rc7

merged at v5.11-rc4, author accept the regression but expect others to fix it

fxmark.hdd_btrfs_MWCL_9_bufferedio.works/sec

[btrfs] 196d59ab9c: -16.4% regression

disk: 1HDD

media: hdd

test: MWCL

fstype: btrfs

directio: bufferedio

cpufreq_governor: performance

lkp-knm01

v5.10-rc7

merged at v5.11-rc1, no response from author yet

reaim.jobs_per_min

[btrfs] 47876f7cef: -7.3% regression

runtime: 300s

nr_task: 100

disk: 1HDD

fs: btrfs

test: disk

cpufreq_governor: performance

lkp-csl-2sp6

v5.10-rc7

merged at v5.11-rc1, no response from author yet

stress-ng.loop.ops_per_sec

[loop] 3b4f85d02a: -42.1% regression

nr_threads: 100%

disk: 1HDD

testtime: 30s

class: device

cpufreq_governor: performance

lkp-ivb-2ep1

v5.10-rc4

merged at v5.11-rc1, the regression fixed in v5.12-rc1

unixbench.score

[locking/rwsem] 617f3ef951: -21.2% regression

runtime: 300s

nr_task: 30%

test: shell8

cpufreq_governor: performance

lkp-cfl-e1

v5.10-rc6

merged at v5.11-rc1, no response from author yet

unixbench.score

[init/console] a91bd6223e: -5.9% regression

runtime: 300s

nr_task: 30%

test: dhry2reg

cpufreq_governor: performance

lkp-cfl-e1

v5.11-rc

merged at v5.11-rc3, author think it’s expected

will-it-scale.per_thread_ops

[sched/hotplug] 2558aacff8: -1.6% regression

nr_task: 100%

mode: thread

test: sched_yield

cpufreq_governor: performance

lkp-cpl-4sp1

v5.10-rc1

merged at v5.11-rc1, author accept the regression and in progress to fix it

fio.write_iops

[crypto] daf88f3757: 5.9% improvement

disk: 2pmem

fs: xfs

mount_option: dax

runtime: 200s

nr_task: 50%

time_based: tb

rw: write

bs: 4k

ioengine: mmap

test_size: 200G

cpufreq_governor: performance

lkp-csl-2sp6

v5.10-rc1

merged at v5.11-rc1

hackbench.throughput

[mm] 3de7d4f25a: 45.2% improvement

nr_threads: 100%

iterations: 4

mode: threads

ipc: socket

cpufreq_governor: performance

lkp-cpl-4sp1

v5.11-rc4

merged at v5.11-rc5

stress-ng.fallocate.ops_per_sec

[btrfs] 639bd575b7: 29.2% improvement

nr_threads: 10%

disk: 1HDD

testtime: 30s

class: filesystem

cpufreq_governor: performance

fs: btrfs

lkp-csl-2sp7

v5.10-rc7

merged at v5.11-rc1

stress-ng.sendfile.ops_per_sec

[mm/filemap.c] 06c0444290: 26.7% improvement

nr_threads: 100%

disk: 1HDD

testtime: 30s

class: pipe

cpufreq_governor: performance

lkp-csl-2sp5

v5.10

merged at v5.11-rc1

stress-ng.vfork.ops_per_sec

[mm, slub] 3286222fc6: 32.1% improvement

nr_threads: 100%

disk: 1HDD

testtime: 60s

sc_pid_max: 4194304

class: scheduler

test: vfork

cpufreq_governor: performance

lkp-csl-2sp7

v5.11-rc7

merged at v5.11

vm-scalability.throughput

[locking/rwsem] 2f06f70292: 385.1% improvement

runtime: 300s

test: small-allocs-mt

cpufreq_governor: performance

lkp-csl-2ap4

v5.10-rc6

merged at v5.11-rc1

will-it-scale.per_process_ops

[driver core]: 2.4 f9aa460672% improvement

nr_task: 16

mode: process

test: eventfd1

cpufreq_governor: performance

lkp-csl-2ap2

v5.10-rc3

merged at v5.11-rc1

will-it-scale.per_thread_ops

[poll] ef0ba05538: 8.9% improvement

nr_task: 50%

mode: thread

test: poll2

cpufreq_governor: performance

lkp-hsw-4ex

v5.11-rc2

merged at v5.11-rc3

 

3.2.fxmark.hdd_btrfs_MWCL_9_bufferedio.works/sec

FxMark is a filesystem benchmark that test multicore scalability.

3.2.1.Scenario: 1HDD-MWCL-btrfs-bufferedio

 

Commit 196d59ab9c was reported to have -16.4% regression of fxmark.hdd_btrfs_MWCL_9_bufferedio.works/sec when comparing to v5.10-rc7. It was merged to mainline at v5.11-rc1.

 

Correlated commits

196d59ab9c

btrfs: switch extent buffer tree lock to rw_semaphore

branch

linus/master

report

[btrfs] 196d59ab9c: -16.4% regression

test scenario

disk: 1HDD

media: hdd

test: MWCL

fstype: btrfs

directio: bufferedio

cpufreq_governor: performance

test machine

lkp-knm01

status

merged at v5.11-rc1

3.3.Unixbench.score

UnixBench is a system benchmark to provide a basic indicator of the performance of a Unix-like system.

3.3.1 scenario: shell8 test

 

 

Commit 617f3ef951 was reported to have -21.2% regression of unixbench.score when comparing to v5.10-rc6. It was merged to mainline at v5.11-rc1.

 

Correlated commits

617f3ef951

locking/rwsem: Remove reader optimistic spinning

branch

linus/master

report

[locking/rwsem] 617f3ef951: -21.2% regression

test scenario

runtime: 300s

nr_task: 30%

test: shell8

cpufreq_governor: performance

test machine

lkp-cfl-e1

status

merged at v5.11-rc1, no response from author yet

3.4.Vm-scalability.throughput

 

vm-scalability exercises functions and regions of the mm subsystem of the Linux kernel. We tested on multiple machines such as HSW EP server, during which we reported improvement on one test scenario.

3.4.1 Scenario: small-allocs-mt test

  

Commit 2f06f70292 was reported to have 385.1% improvement of vm-scalability.throughput when comparing to v5.10-rc6. It was merged to mainline at v5.11-rc1.

 

Correlated commits

2f06f70292

locking/rwsem: Prevent potential lock starvation

branch

linus/master

report

[locking/rwsem] 2f06f70292: 385.1% improvement

test scenario

runtime: 300s

test: small-allocs-mt

cpufreq_governor: performance

test machine

lkp-csl-2ap4

status

merged at v5.11-rc1

 

4Shift-Left Testing

Beyond testing trees in the upstream kernel, 0-Day CI also tests developers’ and maintainers’ trees, which can catch issues earlier and reduce wider impact. We call it “shift-left” testing. During the v5.11 release cycle, 0-Day CI had reported 11 major performance regressions and 9 major improvements by doing shift-left testing. We will share more detailed information together with possible code changes that led to this result for some of these, though the assessment is limited by the test coverage we have now. The whole list is summarized in the report summary section.

4.1Report Summary

0-Day CI had reported 11 performance regressions and 9 improvements by doing shift-left testing on developer and maintainer repos.

 

Test Indicator

Mail

Test Scenario

Test Machine

Status

fsmark.files_per_sec

[btrfs] 53c5e5a1d8: -14.4% regression

iterations: 1x

nr_threads: 1t

disk: 1BRD_48G

fs: btrfs

filesize: 4M

test_size: 24G

sync_method: fsyncBeforeClose

cpufreq_governor: performance

lkp-csl-2ap2

currently not merged, no response from author yet

fsmark.files_per_sec

[blk] 394534b2c5: -44.7% regression

iterations: 1x

nr_threads: 32t

disk: 1SSD

fs: f2fs

filesize: 9B

test_size: 400M

sync_method: fsyncBeforeClose

nr_directories: 16d

nr_files_per_directory: 256fpd

cpufreq_governor: performance

lkp-csl-2sp7

currently not merged, regression has decreased to -3.4% after re-base blk branch, author accpet it and we have some disguession

fxmark.hdd_ext4_no_jnl_MWCL_1_bufferedio.works/sec

[ext4] ef4eebad9c: -9.8% regression

disk: 1HDD

media: hdd

test: MWCL

fstype: ext4_no_jnl

directio: bufferedio

cpufreq_governor: performance

lkp-knm01

currently not merged, no response from author yet

fxmark.hdd_xfs_MWUL_9_directio.works/sec

[xfs] aa0a2ec202: -53.2% regression

disk: 1HDD

media: hdd

test: MWUL

fstype: xfs

directio: directio

cpufreq_governor: performance

lkp-knm01

currently not merged, no response from author yet

hackbench.throughput

[ZEN] 2d56caddc2: -20.5% regression

nr_threads: 100%

iterations: 4

mode: threads

ipc: pipe

cpufreq_governor: performance

lkp-skl-fpga01

currently not merged, no response from author yet

netperf.Throughput_tps

[x86/mce] 7bb39313cd: -4.5% regression

ip: ipv4

runtime: 300s

nr_threads: 16

cluster: cs-localhost

test: TCP_CRR

cpufreq_governor: performance

lkp-csl-2ap3

currently not merged,  but author accept the regression and we have some disguesstion

stress-ng.dentry.ops_per_sec

[xfs] f4da454e15: -36.8% regression

nr_threads: 10%

disk: 1HDD

testtime: 60s

fs: xfs

class: filesystem

test: dentry

cpufreq_governor: performance

lkp-csl-2sp7

currently not merged, no response from author yet

stress-ng.sigrt.ops_per_sec

[sched/fair] 6f6c84fc3b: -2.5% regression

nr_threads: 100%

disk: 1HDD

testtime: 30s

class: interrupt

cpufreq_governor: performance

lkp-ivb-2ep1

currently not merged, no response from author yet

unixbench.score

[entry] 8e01c5f104: -2.2% regression

runtime: 300s

nr_task: 1

test: syscall

cpufreq_governor: performance

lkp-csl-2sp4

currently not merged, no response from author yet

vm-scalability.throughput

[mm/page_alloc.c] 3f595ccb92: -12.0% regression

runtime: 300s

test: lru-file-readtwice

cpufreq_governor: performance

lkp-csl-2ap4

currently not merged, no response from author yet

will-it-scale.per_process_ops

[PCI] dc83615370: -1.2% regression

nr_task: 50%

mode: process

test: mmap1

cpufreq_governor: performance

lkp-skl-fpga01

currently not merged, author is inprogress to investigate

fsmark.files_per_sec

[NFS] 91559d67b2: 48.3% improvement

iterations: 4

disk: 1SSD

nr_threads: 4

fs: ext4

fs2: nfsv4

filesize: 9B

test_size: 10G

sync_method: fsyncBeforeClose

nr_directories: 16d

nr_files_per_directory: 256fpd

cpufreq_governor: performance

lkp-csl-2ap1

currently not merged

hackbench.throughput

[mm] 3de7d4f25a: 45.2% improvement

nr_threads: 100%

iterations: 4

mode: threads

ipc: socket

cpufreq_governor: performance

lkp-cpl-4sp1

merged at v5.11-rc5

netperf.Throughput_tps

[bpf] 8add9a3810: 5.1% improvement

ip: ipv4

runtime: 300s

nr_threads: 1

cluster: cs-localhost

test: UDP_RR

cpufreq_governor: performance

lkp-csl-2ap3

currently not merged

stress-ng.loop.ops_per_sec

[loop] 6cc8e74308: 139.4% improvement

nr_threads: 10%

disk: 1HDD

testtime: 60s

fs: ext4

class: os

test: loop

cpufreq_governor: performance

lkp-csl-2sp5

currently not merged

stress-ng.sendfile.ops_per_sec

[mm/filemap.c] 06c0444290: 26.7% improvement

nr_threads: 100%

disk: 1HDD

testtime: 30s

class: pipe

cpufreq_governor: performance

lkp-csl-2sp5

merged at v5.11-rc1

stress-ng.timerfd.ops_per_sec

[x86/pv] ab234a260b: 6.6% improvement

nr_threads: 10%

disk: 1HDD

testtime: 60s

fs: ext4

class: os

test: timerfd

cpufreq_governor: performance

lkp-csl-2sp5

currently not merged

stress-ng.vfork.ops_per_sec

[mm, slub] 3286222fc6: 32.1% improvement

nr_threads: 100%

disk: 1HDD

testtime: 60s

sc_pid_max: 4194304

class: scheduler

test: vfork

cpufreq_governor: performance

lkp-csl-2sp7

merged at v5.11

vm-scalability.throughput

[mm] 4d8191276e: 43.4% improvement

runtime: 300s

size: 1T

test: lru-shm

cpufreq_governor: performance

lkp-csl-2ap4

currently not merged

will-it-scale.per_thread_ops

[poll] ef0ba05538: 8.9% improvement

nr_task: 50%

mode: thread

test: poll2

cpufreq_governor: performance

lkp-hsw-4ex1

merged at v5.11-rc3

4.2Netperf.Throuhput_tps

Netperf is a benchmark that can be used to measure the performance of many different types of networking. It provides tests for both unidirectional throughput, and end-to-end latency.

4.2.1scenario: TCP_CRR test on localhost

 

Commit 7bb39313cd was reported to have -4.5% regression of netperf.Throughput_tps when comparing to v5.11-rc2.

 

Correlated commits

7bb39313cd

x86/mce: Make mce_timed_out() identify holdout CPUs

branch

tip/ras/core

report

[x86/mce] 7bb39313cd: -4.5% regression

test scenario

ip: ipv4

runtime: 300s

nr_threads: 16

cluster: cs-localhost

test: TCP_CRR

cpufreq_governor: performance

test machine

lkp-csl-2ap3

status

currently not merged,  but author accept the regression and we have some disguesstion

 

4.3stress-ng.dentry.ops_per_sec

stress-ng will stress test a computer system in various selectable ways. It was designed to exercise various physical subsystems of a computer as well as the various operating system kernel interfaces.

4.3.1scenario: dentry test on xfs

 

Commit f4da454e15 was reported to have -36.8% regression of stress-ng.dentry.ops_per_sec when comparing to v5.11-rc7.

 

Correlated commits

f4da454e15

xfs: deferred inode inactivation

branch

djwong-xfs/reserve-rt-metadata-space

report

[xfs] f4da454e15: -36.8% regression

test scenario

nr_threads: 10%

disk: 1HDD

testtime: 60s

fs: xfs

class: filesystem

test: dentry

cpufreq_governor: performance

test machine

lkp-csl-2sp7

status

currently not merged, no response from author yet

4.4vm-scalability

vm-scalability exercises functions and regions of the mm subsystem of the Linux kernel. We tested on multiple machines such as HSW EP server, during which we reported improvement on one test scenario.

4.4.1Scenario: lru-shm test

  

Commit 4d8191276e was reported to have 43.4% improvement of vm-scalability.throughput when comparing to v5.11-rc1

 

Correlated commits

4d8191276e

mm: memcg: add a new MEMCG_UPDATE_BATCH

branch

linux-review/Feng-Tang/mm-page_counter-relayout-structure-to-reduce-false-sharing/20201229-223627

report

[mm] 4d8191276e: 43.4% improvement

test scenario

runtime: 300s

size: 1T

test: lru-shm

cpufreq_governor: performance

test machine

lkp-csl-2ap4

status

currently not merged

5Latest Release Performance Comparing

 

This session gives some information about the performance difference among different kernel releases, especially between v5.11 and v5.10. There are 50+ performance benchmarks running in 0-Day CI, and we selected 9 benchmarks which historically showed the most regressions/improvements reported by 0-Day CI. Some typical configuration/parameters are used to run the test. For some of the regressions from the comparison, 0-Day did not successfully bisect it thus no related report sent out during the release development period, but it is still worth checking. The root cause to cause the regressions won’t be covered in this session.

 

In the following figures, the value on the Y-axis is the relative performance number. We used the v5.10 data as the base (performance number is 100).

5.1test suite: vm-scalability

vm-scalability exercises functions and regions of the mm subsystem of the Linux kernel. Below 2 tests show the typical test results.


vm-scalability Test 1

vm-scalability Test 2

Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                           

 

vm-scalability Test 1

vm-scalability Test 2

test machine

model: Skylake

brand:

cpu_number: 104

memory: 192G

model: Cascade Lake

brand: Intel(R) Xeon(R) Gold 6252 CPU @ 2.10GHz

cpu_number: 96

memory: 512G

runtime

300s

600s

size

1T

4T

vm-scalability test parameter

test case: lru-shm
test case: anon-cow-seq

 

performance summary

vm-scalability.throughput on kernel v5.11 has 4.74% improvement when comparing to v5.10

vm-scalability.throughput on kernel v5.11 has -39.74% regression when comparing to v5.10

 

5.2test suite: will-it-scale

Will-it-scale takes a test case and runs it from 1 through to n parallel copies to see if the test case will scale. It builds both process and threads based tests in order to see any differences between the two.


will-it-scale  Test 1

Will-it-scale Test 2

Here are the parameters and performance test summary for above tests:                                                                                                                                                                                                                          

 

will-it-scale Test 1

will-it-scale Test 2

test machine

model: Cascade Lake

brand: Intel(R) Xeon(R) Platinum 9242 CPU @ 2.30GHz

cpu_number: 192

memory: 192G

model: Knights Mill

brand: Intel(R) Xeon Phi(TM) CPU 7295 @ 1.50GHz

cpu_number: 288

memory: 80G

nr_task

No requirement

192

will-it-scale test parameter

mode: process
test: eventfd1

mode: process

test: read2

summary

will-it-scale.per_process_ops on kernel v5.11 has 3.83% improvement when comparing to v5.10

will-it-scale.per_process_ops on kernel v5.11 has -5.5% regression when comparing to v5.10

 

5.3test suite: unixbench

UnixBench is a system benchmark to provide a basic indicator of the performance of a Unix-like system.


Unixbench Test 1

Unixbench Test 2

Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                           

 

Unixbench Test 1

Unixbench Test 2

test machine

model: Cascade Lake

brand: Intel(R) Xeon(R) CPU @ 2.30GHz

cpu_number: 96

memory: 128G

model: Coffee Lake

brand: Intel(R) Xeon(R) E-2278G CPU @ 3.40GHz

cpu_number: 16

memory: 32G

runtime

300s

300s

nr_task

1

30%

unixbench test parameter

test: syscall

test case: shell8

 

performance summary

unixbench.score on kernel v5.11 has 5.04% improvement when comparing to v5.10

unixbench.score on kernel v5.11 has -23.69% regression when comparing to v5.10

 

5.4test suite: reaim

reaim updates and improves the existing Open Source AIM 7 benchmark. aim7 is a traditional UNIX system level benchmark suite which is used to test and measure the performance of a multiuser system.


reaim Test 1                             

reaim Test 2

Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                                      

 

reaim Test 1

reaim Test 2

test machine

model: Cooper Lake

brand: Intel(R) Xeon(R) Gold 5318H CPU @ 2.50GHz

cpu_number: 144

memory: 128G

model: Cascade Lake

brand: Intel(R) Xeon(R) Gold 6252 CPU @ 2.10GHz

cpu_number: 96

memory: 256G

runtime

300s

300s

nr_task

100%

100%

disk

No requirement

1HDD

fs

No requirement

btrfs

 

reaim test parameter

test: short

test case: disk

performance  summary

reaim.jobs_per_min_child on kernel v5.11 has 27.7% undefined when comparing to v5.10

reaim.jobs_per_min on kernel v5.11 has -12.06% regression when comparing to v5.10

 

 

5.5test suite: pigz

pigz, which stands for Parallel Implementation of GZip, is a fully functional replacement for gzip that exploits multiple processors and multiple cores to the hilt when compressing data.

pigz Test 1

Here are the test configuration and performance test summary for above tests:         

 

pigz Test 1

test machine

model: Knights Mill

brand: Intel(R) Xeon Phi(TM) CPU 7295 @ 1.50GHz

cpu_number: 288

memory: 80G

nr_threads

100%

pigz Test parameter

blocksize: 512K

 

performance  summary

pigz.throughput on kernel v5.11 has 16.71% improvement when comparing to v5.10

 

           

5.6test suite: netperf

Netperf is a benchmark that can be used to measure the performance of many different types of networking. It provides tests for both unidirectional throughput, and end-to-end latency.


netperf Test 1

netperf Test 2

Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                                          

 

netperf Test 1

netperf Test 2

test machine

model: Cooper Lake

brand: Intel(R) Xeon(R) Gold 5318H CPU @ 2.50GHz

cpu_number: 144

memory: 128G

model: Cooper Lake

brand: Intel(R) Xeon(R) Gold 5318H CPU @ 2.50GHz

cpu_number: 144

memory: 128G

disable_latency_stats

1

1

set_nic_irq_affinity

1

1

runtime

300s

900s

nr_threads

1

50%

ip

ipv4

Ipv4

netperf test parameter

test case: UDP_RR

 

test case: TCP_STREAM

 

performance  summary

netperf.Throughput_tps on kernel v5.11 has 3.63% improvement when comparing to v5.10

netperf.Throughput_tps on kernel v5.11 has 6.14% regression when comparing to v5.10

 

5.7test suite: hackbench

Hackbench is both a benchmark and a stress test for the Linux kernel scheduler. It's  main job  is  to  create a specified number of pairs of schedulable entities (either threads or traditional processes) which communicate via either sockets or pipes and time how long  it takes for each pair to send data back and forth.

hackbench Test 1

 
Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                           

 

hackbench Test 1

test machine

model: Skylake

brand:

cpu_number: 104

memory: 192G

disable_latency_stats

1

nr_task

100%

unixbench test parameter

mode: threads

ipc: socket

performance summary

hackbench.throughput on kernel v5.11 has 27.25% improvement when comparing to v5.10

 

           

5.8test suite: fio

Fio was originally written to save me the hassle of writing special test case programs when I wanted to test a specific workload, either for performance reasons or to find/reproduce a bug.

 

 

fio Test 1 


 

Here are the test configuration and performance test summary for above tests:                                                                                                                                                                                                                          

 

fio Test 1

test machine

model: Cascade Lake

brand: Intel(R) Xeon(R) CPU @ 2.20GHz

cpu_number: 192

memory: 192G

runtime

300s

file system

xfs

disk

1SSD

boot_params

No requirement

nr_task

32

time_based

No requirement

fio test parameter

fio-setup-basic:

  rw: randwrite

  bs: 4k

  ioengine: sync

  test_size: 256g

performance  summary

fio.write_iops on kernel v5.11 has -26.82% regression when comparing to v5.10

5.9test suite: ebizzy

ebizzy is designed to generate a workload resembling common web application server workloads. It is highly threaded, has a large in-memory working set, and allocates and deallocates memory frequently.


                                      ebizzy Test 1

ebizzy Test 2

Here are the test configuration and performance test summary for above test:                                                                                                                                                                                                                          

 

ebizzy Test 1

ebizzy Test 2

test machine

model: Knights Mill

brand: Intel(R) Xeon Phi(TM) CPU 7295 @ 1.50GHz

cpu_number: 288

memory: 80G

model: Knights Mill

brand: Intel(R) Xeon Phi(TM) CPU 7295 @ 1.50GHz

cpu_number: 288

memory: 80G

transparent_hugepage

No requirement

No requirement

nr_threads

200%

200%

iterations

100x

100x

ebizzy test parameter

duration: 10s

duration: 10s

performance  summary

ebizzy.throughput on kernel v5.11 has 53.7% improvement when comparing to v5.10

ebizzy.throughput.per_thread.stddev_percent on kernel v5.11 has -34.63% undefined when comparing to v5.10

 

6Test Machines

6.1IVB Desktop

model

Ivy Bridge

brand

Intel(R) Core(TM) i3-3220 CPU @ 3.30GHz

cpu number

8

memory

16G

 

model

Ivy Bridge

brand

Intel(R) Core(TM) i3-3220 CPU @ 3.30GHz

cpu number

4

memory

8G

 

6.2SKL SP

model

Skylake

brand

Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz

cpu number

80

memory

64G

 

6.3BDW EP

model

Broadwell-EP

brand

Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz

cpu number

88

memory

128G

 

6.4HSW EP

model

Haswell-EP

brand

Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz

cpu number

72

memory

128G

 

6.5IVB EP

model

Ivy Bridge-EP

brand

Intel(R) Xeon(R) CPU E5-2690 v2 @ 3.00GHz

cpu number

40

memory

384G

 

model

Ivytown Ivy Bridge-EP

brand

Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz

cpu number

48

memory

64G

 

6.6HSX EX

model

Brickland Haswell-EX

brand

Intel(R) Xeon(R) CPU E7-8890 v3 @ 2.50GHz

cpu number

144

memory

512G