ちゃんるいすのブログ

オタクエンジニアの雑記

AWS EC2 m6g, c6g, r6g は本当に速くなったのか


m系だけで比較 UnixBench

AMI
m6g.large: amzn2-ami-hvm-2.0.20200406.0-arm64-gp2
m5.large: amzn2-ami-hvm-2.0.20200406.0-x86_64-gp2

# yum update -y
# yum groupinstall '@Development Tools'
# wget https://github.com/kdlucas/byte-unixbench/archive/master.zip
# unzip master.zip

m5.large

2 CPUs in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       38850809.4 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     4363.1 MWIPS (9.1 s, 7 samples)
Execl Throughput                               4709.3 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        594108.7 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          154550.5 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       1970248.9 KBps  (30.0 s, 2 samples)
Pipe Throughput                              747070.6 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                  66250.9 lps   (10.0 s, 7 samples)
Process Creation                              11919.4 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   7624.2 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1214.2 lpm   (60.0 s, 2 samples)
System Call Overhead                         412715.6 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   38850809.4   3329.1
Double-Precision Whetstone                       55.0       4363.1    793.3
Execl Throughput                                 43.0       4709.3   1095.2
File Copy 1024 bufsize 2000 maxblocks          3960.0     594108.7   1500.3
File Copy 256 bufsize 500 maxblocks            1655.0     154550.5    933.8
File Copy 4096 bufsize 8000 maxblocks          5800.0    1970248.9   3397.0
Pipe Throughput                               12440.0     747070.6    600.5
Pipe-based Context Switching                   4000.0      66250.9    165.6
Process Creation                                126.0      11919.4    946.0
Shell Scripts (1 concurrent)                     42.4       7624.2   1798.2
Shell Scripts (8 concurrent)                      6.0       1214.2   2023.7
System Call Overhead                          15000.0     412715.6    275.1
                                                                   ========
System Benchmarks Index Score                                        1021.9

------------------------------------------------------------------------
Benchmark Run: 月  5月 18 2020 09:31:17 - 09:59:20
2 CPUs in system; running 2 parallel copies of tests

Dhrystone 2 using register variables       55968401.9 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     7327.2 MWIPS (9.3 s, 7 samples)
Execl Throughput                               6701.5 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        789430.5 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          204553.5 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2679537.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                             1012333.1 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 282192.2 lps   (10.0 s, 7 samples)
Process Creation                              19656.0 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   8858.5 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1224.7 lpm   (60.0 s, 2 samples)
System Call Overhead                         540338.4 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   55968401.9   4795.9
Double-Precision Whetstone                       55.0       7327.2   1332.2
Execl Throughput                                 43.0       6701.5   1558.5
File Copy 1024 bufsize 2000 maxblocks          3960.0     789430.5   1993.5
File Copy 256 bufsize 500 maxblocks            1655.0     204553.5   1236.0
File Copy 4096 bufsize 8000 maxblocks          5800.0    2679537.8   4619.9
Pipe Throughput                               12440.0    1012333.1    813.8
Pipe-based Context Switching                   4000.0     282192.2    705.5
Process Creation                                126.0      19656.0   1560.0
Shell Scripts (1 concurrent)                     42.4       8858.5   2089.3
Shell Scripts (8 concurrent)                      6.0       1224.7   2041.2
System Call Overhead                          15000.0     540338.4    360.2
                                                                   ========
System Benchmarks Index Score                                        1523.2

m6g.large

2 CPUs in system; running 1 parallel copy of tests

Dhrystone 2 using register variables       40582507.5 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     5912.0 MWIPS (9.6 s, 7 samples)
Execl Throughput                               7006.5 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1041850.6 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          291910.6 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2952450.1 KBps  (30.0 s, 2 samples)
Pipe Throughput                             1790930.7 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 134169.8 lps   (10.0 s, 7 samples)
Process Creation                              11434.1 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   8528.6 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1658.2 lpm   (60.0 s, 2 samples)
System Call Overhead                        1733100.7 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   40582507.5   3477.5
Double-Precision Whetstone                       55.0       5912.0   1074.9
Execl Throughput                                 43.0       7006.5   1629.4
File Copy 1024 bufsize 2000 maxblocks          3960.0    1041850.6   2630.9
File Copy 256 bufsize 500 maxblocks            1655.0     291910.6   1763.8
File Copy 4096 bufsize 8000 maxblocks          5800.0    2952450.1   5090.4
Pipe Throughput                               12440.0    1790930.7   1439.7
Pipe-based Context Switching                   4000.0     134169.8    335.4
Process Creation                                126.0      11434.1    907.5
Shell Scripts (1 concurrent)                     42.4       8528.6   2011.5
Shell Scripts (8 concurrent)                      6.0       1658.2   2763.6
System Call Overhead                          15000.0    1733100.7   1155.4
                                                                   ========
System Benchmarks Index Score                                        1649.2

------------------------------------------------------------------------
Benchmark Run: 月  5月 18 2020 09:35:22 - 10:03:18
2 CPUs in system; running 2 parallel copies of tests

Dhrystone 2 using register variables       81093723.6 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                    11816.9 MWIPS (9.6 s, 7 samples)
Execl Throughput                              11173.2 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1311901.0 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          405826.8 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       3190465.5 KBps  (30.0 s, 2 samples)
Pipe Throughput                             3586570.9 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 638429.7 lps   (10.0 s, 7 samples)
Process Creation                              22963.7 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                  12549.6 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1706.7 lpm   (60.0 s, 2 samples)
System Call Overhead                        2563990.0 lps   (10.0 s, 7 samples)

System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   81093723.6   6948.9
Double-Precision Whetstone                       55.0      11816.9   2148.5
Execl Throughput                                 43.0      11173.2   2598.4
File Copy 1024 bufsize 2000 maxblocks          3960.0    1311901.0   3312.9
File Copy 256 bufsize 500 maxblocks            1655.0     405826.8   2452.1
File Copy 4096 bufsize 8000 maxblocks          5800.0    3190465.5   5500.8
Pipe Throughput                               12440.0    3586570.9   2883.1
Pipe-based Context Switching                   4000.0     638429.7   1596.1
Process Creation                                126.0      22963.7   1822.5
Shell Scripts (1 concurrent)                     42.4      12549.6   2959.8
Shell Scripts (8 concurrent)                      6.0       1706.7   2844.5
System Call Overhead                          15000.0    2563990.0   1709.3
                                                                   ========
System Benchmarks Index Score                                        2775.8

価格も m5.large より安いしパフォーマンスも良いなら、m6g 使います(ARM ベースで問題ないなら)