[GCP] Cloud Shellのベンチマーク

Google Cloud Platformに関する作業などがブラウザから簡単に行うことができるCloud Shell
外出先等でちょっと用事があるときに非常に便利ですよね。
今回はこのCloud ShellのベンチマークをUnixBenchで採ってみました。
ところで、Cloud ShellはGCEのg1-smallらしいです。(ブースト時はn1-standard-1)
前はg1-smallではなくf1-microだった気がするんですけど、気のせいですかね?

特徴

・Compute Engine での一時的な仮想マシン インスタンス
・ウェブブラウザからインスタンスへのコマンドライン アクセス
・組み込みのコードエディタ ベータ版
・5 GB の永続ディスク ストレージ
・プレインストールされた Google Cloud SDK とその他のツール
・Java、Go、Python、Node.js、PHP、Ruby、.NET への言語対応
・ウェブ プレビュー機能
・GCP Console のプロジェクトとリソースに対するアクセスのビルトイン承認
https://cloud.google.com/shell/docs/features

通常はCloud Shellを開いた時、g1-smallがプロビジョニングされます。
ブーストモードを有効にした場合はn1-standard-1としてプロビジョニングされます。

また、実は週単位で使用できる時間に制限があるようです。

サーバースペック

g1-small

0.5 基の仮想 CPU と 1.70 GB のメモリを備え、共有物理コアを基盤とした共有コア マシンタイプ。
https://cloud.google.com/compute/docs/machine-types

:~$ free -m
              total        used        free      shared  buff/cache   available
Mem:           1695         362         463           0         869        1180
Swap:             0           0           0

:~$ df -Th
Filesystem     Type     Size  Used Avail Use% Mounted on
overlay        overlay   31G   22G  8.4G  73% /
tmpfs          tmpfs    848M     0  848M   0% /dev
tmpfs          tmpfs    848M     0  848M   0% /sys/fs/cgroup
/dev/sda1      ext4      31G   22G  8.4G  73% /root
/dev/sdb1      ext4     4.8G   12M  4.6G   1% /home
overlayfs      overlay  1.0M  108K  916K  11% /etc/ssh/keys
tmpfs          tmpfs    848M  720K  847M   1% /run/metrics
shm            tmpfs     64M     0   64M   0% /dev/shm
overlayfs      overlay  1.0M  108K  916K  11% /etc/ssh/ssh_host_dsa_key
tmpfs          tmpfs    848M     0  848M   0% /run/google/devshell

n1-standard-1

1 基の仮想 CPU と 3.75 GB のメモリを備えた標準マシンタイプ。
https://cloud.google.com/compute/docs/machine-types

~$ free -m
              total        used        free      shared  buff/cache   available
Mem:           3698         352        2171           0        1175        3119
Swap:             0           0           0

:~$ df -Th
Filesystem     Type     Size  Used Avail Use% Mounted on
overlay        overlay   31G   23G  8.3G  73% /
tmpfs          tmpfs    1.9G     0  1.9G   0% /dev
tmpfs          tmpfs    1.9G     0  1.9G   0% /sys/fs/cgroup
/dev/sda1      ext4      31G   23G  8.3G  73% /root
/dev/sdb1      ext4     4.8G   12M  4.6G   1% /home
overlayfs      overlay  1.0M  108K  916K  11% /etc/ssh/keys
tmpfs          tmpfs    1.9G  676K  1.9G   1% /run/metrics
shm            tmpfs     64M     0   64M   0% /dev/shm
overlayfs      overlay  1.0M  108K  916K  11% /etc/ssh/ssh_host_rsa_key
tmpfs          tmpfs    1.9G     0  1.9G   0% /run/google/devshell

$ cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU @ 2.50GHz
stepping        : 4
microcode       : 0x1
cpu MHz         : 2500.000
cache size      : 30720 KB
physical id     : 0
siblings        : 1
core id         : 0
cpu cores       : 1
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 x2apic popcnt aes xsave avx f16c rdrand hypervisor lahf_lm pti fsgsbase tsc_adjust smep erms xsaveopt
bugs            : cpu_meltdown spectre_v1 spectre_v2 l1tf
bogomips        : 5000.00
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

UnixBench

通常時

通常時のUnixBench。

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: cs-6000-devshell-vm-xxxxxx: GNU/Linux
   OS: GNU/Linux -- 4.14.33+ -- #1 SMP Sat Aug 11 08:05:16 PDT 2018
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   CPU 0: Intel(R) Xeon(R) CPU @ 2.50GHz (5000.0 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   00:35:02 up 15 min,  2 users,  load average: 0.14, 0.33, 0.20; runlevel

------------------------------------------------------------------------
Benchmark Run: Mon Aug 20 2018 00:35:02 - 01:04:05
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       29138732.1 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     2942.8 MWIPS (13.0 s, 7 samples)
Execl Throughput                               1947.3 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        200275.9 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           56955.0 KBps  (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        538326.0 KBps  (30.0 s, 2 samples)
Pipe Throughput                              310676.5 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                  79299.3 lps   (10.0 s, 7 samples)
Process Creation                               5724.0 lps   (30.1 s, 2 samples)
Shell Scripts (1 concurrent)                   3252.4 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    397.4 lpm   (60.2 s, 2 samples)
System Call Overhead                         180235.1 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   29138732.1   2496.9
Double-Precision Whetstone                       55.0       2942.8    535.1
Execl Throughput                                 43.0       1947.3    452.9
File Copy 1024 bufsize 2000 maxblocks          3960.0     200275.9    505.7
File Copy 256 bufsize 500 maxblocks            1655.0      56955.0    344.1
File Copy 4096 bufsize 8000 maxblocks          5800.0     538326.0    928.1
Pipe Throughput                               12440.0     310676.5    249.7
Pipe-based Context Switching                   4000.0      79299.3    198.2
Process Creation                                126.0       5724.0    454.3
Shell Scripts (1 concurrent)                     42.4       3252.4    767.1
Shell Scripts (8 concurrent)                      6.0        397.4    662.3
System Call Overhead                          15000.0     180235.1    120.2
                                                                   ========
System Benchmarks Index Score                                         475.7

ブーストモード

ブーストモードを有効にしたときのUnixBench。

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: cs-6000-devshell-vm-xxxxx: GNU/Linux
   OS: GNU/Linux -- 4.14.33+ -- #1 SMP Sat Aug 11 08:05:16 PDT 2018
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   CPU 0: Intel(R) Xeon(R) CPU @ 2.50GHz (5000.0 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   01:12:24 up 10:31,  2 users,  load average: 0.26, 0.10, 0.03; runlevel

------------------------------------------------------------------------
Benchmark Run: Mon Aug 20 2018 01:12:24 - 01:40:32
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       30024941.0 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     3797.8 MWIPS (9.8 s, 7 samples)
Execl Throughput                               3630.8 lps   (29.5 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        332742.6 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           90811.0 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        903244.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                              487114.9 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 115958.0 lps   (10.0 s, 7 samples)
Process Creation                              10922.9 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   6742.0 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    838.9 lpm   (60.0 s, 2 samples)
System Call Overhead                         276681.1 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   30024941.0   2572.8
Double-Precision Whetstone                       55.0       3797.8    690.5
Execl Throughput                                 43.0       3630.8    844.4
File Copy 1024 bufsize 2000 maxblocks          3960.0     332742.6    840.3
File Copy 256 bufsize 500 maxblocks            1655.0      90811.0    548.7
File Copy 4096 bufsize 8000 maxblocks          5800.0     903244.8   1557.3
Pipe Throughput                               12440.0     487114.9    391.6
Pipe-based Context Switching                   4000.0     115958.0    289.9
Process Creation                                126.0      10922.9    866.9
Shell Scripts (1 concurrent)                     42.4       6742.0   1590.1
Shell Scripts (8 concurrent)                      6.0        838.9   1398.1
System Call Overhead                          15000.0     276681.1    184.5
                                                                   ========
System Benchmarks Index Score                                         770.0

Index Score

通常時とブーストモードで3回UnixBenchを走らせた結果デス。

通常時ブーストモードg1-smalln1-standard-1f1-micro
1475.7770.0847.71313.6152.1
2483.0768.8
3476.8769.0

おまけ

比較のため、f1-micro, g1-small, n1-standard-1のUnixBenchも採ったので載せておきます。

g1-small
========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: g1small: GNU/Linux
   OS: GNU/Linux -- 4.9.0-7-amd64 -- #1 SMP Debian 4.9.110-3+deb9u2 (2018-08-13)
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
   CPU 0: Intel(R) Xeon(R) CPU @ 2.30GHz (4600.0 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   01:53:20 up 3 min,  1 user,  load average: 0.08, 0.04, 0.01; runlevel Aug

------------------------------------------------------------------------
Benchmark Run: Mon Aug 20 2018 01:53:20 - 02:22:12
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       30951363.0 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     3507.5 MWIPS (12.2 s, 7 samples)
Execl Throughput                               2996.6 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        458268.7 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          133952.0 KBps  (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       1218898.4 KBps  (30.0 s, 2 samples)
Pipe Throughput                              776309.5 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 135166.1 lps   (10.0 s, 7 samples)
Process Creation                               8439.8 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   4922.6 lpm   (60.1 s, 2 samples)
Shell Scripts (8 concurrent)                    617.5 lpm   (60.1 s, 2 samples)
System Call Overhead                         527722.7 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   30951363.0   2652.2
Double-Precision Whetstone                       55.0       3507.5    637.7
Execl Throughput                                 43.0       2996.6    696.9
File Copy 1024 bufsize 2000 maxblocks          3960.0     458268.7   1157.2
File Copy 256 bufsize 500 maxblocks            1655.0     133952.0    809.4
File Copy 4096 bufsize 8000 maxblocks          5800.0    1218898.4   2101.5
Pipe Throughput                               12440.0     776309.5    624.0
Pipe-based Context Switching                   4000.0     135166.1    337.9
Process Creation                                126.0       8439.8    669.8
Shell Scripts (1 concurrent)                     42.4       4922.6   1161.0
Shell Scripts (8 concurrent)                      6.0        617.5   1029.2
System Call Overhead                          15000.0     527722.7    351.8
                                                                   ========
System Benchmarks Index Score                                         847.7
n1-standard-1
========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: n1standard1: GNU/Linux
   OS: GNU/Linux -- 4.9.0-7-amd64 -- #1 SMP Debian 4.9.110-3+deb9u2 (2018-08-13)
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
   CPU 0: Intel(R) Xeon(R) CPU @ 2.30GHz (4600.0 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   01:53:04 up 3 min,  1 user,  load average: 0.19, 0.07, 0.02; runlevel Aug

------------------------------------------------------------------------
Benchmark Run: Mon Aug 20 2018 01:53:04 - 02:21:11
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       31460987.6 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     4176.3 MWIPS (9.9 s, 7 samples)
Execl Throughput                               5369.1 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        735957.9 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          209307.4 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       1854594.5 KBps  (30.0 s, 2 samples)
Pipe Throughput                             1130884.6 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 208971.2 lps   (10.0 s, 7 samples)
Process Creation                              15182.0 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   9874.3 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1206.0 lpm   (60.0 s, 2 samples)
System Call Overhead                         769258.9 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   31460987.6   2695.9
Double-Precision Whetstone                       55.0       4176.3    759.3
Execl Throughput                                 43.0       5369.1   1248.6
File Copy 1024 bufsize 2000 maxblocks          3960.0     735957.9   1858.5
File Copy 256 bufsize 500 maxblocks            1655.0     209307.4   1264.7
File Copy 4096 bufsize 8000 maxblocks          5800.0    1854594.5   3197.6
Pipe Throughput                               12440.0    1130884.6    909.1
Pipe-based Context Switching                   4000.0     208971.2    522.4
Process Creation                                126.0      15182.0   1204.9
Shell Scripts (1 concurrent)                     42.4       9874.3   2328.8
Shell Scripts (8 concurrent)                      6.0       1206.0   2010.1
System Call Overhead                          15000.0     769258.9    512.8
                                                                   ========
System Benchmarks Index Score                                        1313.6
f1-micro
========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)

   System: f1micro: GNU/Linux
   OS: GNU/Linux -- 4.9.0-7-amd64 -- #1 SMP Debian 4.9.110-3+deb9u2 (2018-08-13)
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
   CPU 0: Intel(R) Xeon(R) CPU @ 2.30GHz (4600.0 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   01:52:55 up 2 min,  1 user,  load average: 0.22, 0.10, 0.03; runlevel Aug

------------------------------------------------------------------------
Benchmark Run: Mon Aug 20 2018 01:52:55 - 02:23:43
1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       17570154.4 lps   (10.1 s, 7 samples)
Double-Precision Whetstone                      838.1 MWIPS (28.3 s, 7 samples)
Execl Throughput                               1218.6 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        161377.9 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           39812.8 KBps  (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        330861.9 KBps  (30.0 s, 2 samples)
Pipe Throughput                              340018.4 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                  62298.5 lps   (10.0 s, 7 samples)
Process Creation                               3822.6 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   2085.7 lpm   (60.1 s, 2 samples)
Shell Scripts (8 concurrent)                    288.9 lpm   (60.2 s, 2 samples)
System Call Overhead                         228185.5 lps   (10.1 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   17570154.4   1505.6
Double-Precision Whetstone                       55.0        838.1    152.4
Execl Throughput                                 43.0       1218.6    283.4
File Copy 1024 bufsize 2000 maxblocks          3960.0     161377.9    407.5
File Copy 256 bufsize 500 maxblocks            1655.0      39812.8    240.6
File Copy 4096 bufsize 8000 maxblocks          5800.0     330861.9    570.5
Pipe Throughput                               12440.0     340018.4    273.3
Pipe-based Context Switching                   4000.0      62298.5    155.7
Process Creation                                126.0       3822.6    303.4
Shell Scripts (1 concurrent)                     42.4       2085.7    491.9
Shell Scripts (8 concurrent)                      6.0        288.9    481.4
System Call Overhead                          15000.0     228185.5    152.1
                                                                   ========
System Benchmarks Index Score                                         330.4

最後に

Index Scoreをざっと見ると、Cloud Shellはg1-smallがプロビジョニングされているとは言えだいぶ制限されているように見えます。
通常時はf1-microより少しいいぐらいで、ブースト時にg1-small相当のような感じでしょうか。
Cloud Shellは無料で使えるので勿論十分すぎるといえば十分すぎますね。

SNSでもご購読できます。

コメントを残す