NIFCLOUDの新しいサーバータイプ Type-cが2019/4/23にリリースされました。
Type-eよりもCPUの性能に制限が掛かったTypeのようなので、測定してみました。
SYSBENCH(CPU)
Type-h small
CPU speed:
events per second: 1051.45
General statistics:
total time: 10.0004s
total number of events: 10517
Latency (ms):
min: 0.93
avg: 0.95
max: 1.37
95th percentile: 1.01
sum: 9992.60
Threads fairness:
events (avg/stddev): 10517.0000/0.00
execution time (avg/stddev): 9.9926/0.00
Type-e e-small
CPU speed:
events per second: 885.15
General statistics:
total time: 10.1905s
total number of events: 9022
Latency (ms):
min: 0.93
avg: 1.13
max: 508.50
95th percentile: 1.03
sum: 10182.41
Threads fairness:
events (avg/stddev): 9022.0000/0.00
execution time (avg/stddev): 10.1824/0.00
Type-c c-small
CPU speed:
events per second: 830.27
General statistics:
total time: 10.3588s
total number of events: 8602
Latency (ms):
min: 0.97
avg: 1.20
max: 421.21
95th percentile: 1.01
sum: 10354.67
Threads fairness:
events (avg/stddev): 8602.0000/0.00
execution time (avg/stddev): 10.3547/0.00
評価
Type-hと比べて、Type-eは15%減、Type-cは18%減に見えますね。
UNIXBENCH
Type-h small
1 CPU in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 37408190.4 lps (10.0 s, 7 samples)
Double-Precision Whetstone 3297.7 MWIPS (16.6 s, 7 samples)
Execl Throughput 3680.0 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 751840.6 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 199694.0 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 2365735.7 KBps (30.0 s, 2 samples)
Pipe Throughput 1046250.0 lps (10.0 s, 7 samples)
Pipe-based Context Switching 209100.6 lps (10.0 s, 7 samples)
Process Creation 11479.7 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 5521.6 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 756.9 lpm (60.0 s, 2 samples)
System Call Overhead 1020147.1 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 37408190.4 3205.5
Double-Precision Whetstone 55.0 3297.7 599.6
Execl Throughput 43.0 3680.0 855.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 751840.6 1898.6
File Copy 256 bufsize 500 maxblocks 1655.0 199694.0 1206.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 2365735.7 4078.9
Pipe Throughput 12440.0 1046250.0 841.0
Pipe-based Context Switching 4000.0 209100.6 522.8
Process Creation 126.0 11479.7 911.1
Shell Scripts (1 concurrent) 42.4 5521.6 1302.3
Shell Scripts (8 concurrent) 6.0 756.9 1261.5
System Call Overhead 15000.0 1020147.1 680.1
========
System Benchmarks Index Score 1174.3
Type-e e-small
1 CPU in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 24572314.7 lps (10.1 s, 7 samples)
Double-Precision Whetstone 2805.8 MWIPS (9.7 s, 7 samples)
Execl Throughput 1784.0 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 419723.8 KBps (30.1 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 112393.0 KBps (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 1317388.7 KBps (30.0 s, 2 samples)
Pipe Throughput 642509.6 lps (10.1 s, 7 samples)
Pipe-based Context Switching 112149.3 lps (10.1 s, 7 samples)
Process Creation 5857.8 lps (30.2 s, 2 samples)
Shell Scripts (1 concurrent) 2760.4 lpm (60.1 s, 2 samples)
Shell Scripts (8 concurrent) 370.6 lpm (60.1 s, 2 samples)
System Call Overhead 606825.9 lps (10.1 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 24572314.7 2105.6
Double-Precision Whetstone 55.0 2805.8 510.1
Execl Throughput 43.0 1784.0 414.9
File Copy 1024 bufsize 2000 maxblocks 3960.0 419723.8 1059.9
File Copy 256 bufsize 500 maxblocks 1655.0 112393.0 679.1
File Copy 4096 bufsize 8000 maxblocks 5800.0 1317388.7 2271.4
Pipe Throughput 12440.0 642509.6 516.5
Pipe-based Context Switching 4000.0 112149.3 280.4
Process Creation 126.0 5857.8 464.9
Shell Scripts (1 concurrent) 42.4 2760.4 651.0
Shell Scripts (8 concurrent) 6.0 370.6 617.6
System Call Overhead 15000.0 606825.9 404.6
========
System Benchmarks Index Score 668.6
Type-c c-small
1 CPU in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 24675144.6 lps (10.1 s, 7 samples)
Double-Precision Whetstone 1691.2 MWIPS (32.9 s, 7 samples)
Execl Throughput 1939.6 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 424038.1 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 115358.1 KBps (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 1305745.8 KBps (30.0 s, 2 samples)
Pipe Throughput 660825.0 lps (10.0 s, 7 samples)
Pipe-based Context Switching 120396.5 lps (10.0 s, 7 samples)
Process Creation 5750.1 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 2905.0 lpm (60.1 s, 2 samples)
Shell Scripts (8 concurrent) 398.0 lpm (60.1 s, 2 samples)
System Call Overhead 637112.7 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 24675144.6 2114.4
Double-Precision Whetstone 55.0 1691.2 307.5
Execl Throughput 43.0 1939.6 451.1
File Copy 1024 bufsize 2000 maxblocks 3960.0 424038.1 1070.8
File Copy 256 bufsize 500 maxblocks 1655.0 115358.1 697.0
File Copy 4096 bufsize 8000 maxblocks 5800.0 1305745.8 2251.3
Pipe Throughput 12440.0 660825.0 531.2
Pipe-based Context Switching 4000.0 120396.5 301.0
Process Creation 126.0 5750.1 456.4
Shell Scripts (1 concurrent) 42.4 2905.0 685.1
Shell Scripts (8 concurrent) 6.0 398.0 663.4
System Call Overhead 15000.0 637112.7 424.7
========
System Benchmarks Index Score 660.9
評価
hはもちろんスコア高いんですが、eとcにそれほど差違がない…。
整数演算はさほど差がなく浮動小数点演算だけ大差ついてるのでその辺絞ってるのかもですね。
iperf
Type-h small
Interval Transfer Bandwidth
0.0-10.0 sec 10.1 GBytes 8.65 Gbits/sec
Type-e e-small
Interval Transfer Bandwidth
0.0-10.0 sec 240 MBytes 201 Mbits/sec
Type-e c-small
Interval Transfer Bandwidth
0.0-10.0 sec 120 MBytes 101 Mbits/sec
評価
hは特に帯域制限されてなさそうですが、eは200Mbps、cは100Mbps制限ですね。
総合評価
Type-cはType-eから帯域制限を更に厳しくしたプランって感じですね。CPUは若干下げた設定になってるのかもですね。