m1.small c1.medium m1.large m1.xlarge c1.xlarge Dhrystone 2 using register variables 318.2 839 958.8 961.4 1109 Double-Precision Whetstone 396.4 376.5 301.6 394.7 487.2 Execl Throughput 196.7 371.7 - - - File Copy 1024 bufsize 2000 maxblocks 113.8 354.5 537.8 562.6 643.4 File Copy 256 bufsize 500 maxblocks 74.9 221.3 348.5 361.1 415.8 File Copy 4096 bufsize 8000 maxblocks 239.5 739.1 906 920.4 1101.7 Pipe Throughput 48 144.9 259 270.3 314.1 Pipe-based Context Switching 55 89.6 199.8 156.9 217 Process Creation 111.1 269.9 272.2 281.1 320 Shell Scripts (8 concurrent) 352.5 1076.5 1310.3 1637.7 1227.8 System Call Overhead 447.2 551 226.8 243.7 283.4
m2.xlarge m2.2xlarge m2.4xlarge cc1.4xlarge Dhrystone 2 using register variables 1311.4 1313.4 1313 1438.6 Double-Precision Whetstone 549.9 550.5 550.8 604 Execl Throughput - - - - File Copy 1024 bufsize 2000 maxblocks 749.2 743.5 741.4 1645.2 File Copy 256 bufsize 500 maxblocks 478.5 478.2 471.1 1049.4 File Copy 4096 bufsize 8000 maxblocks 1512.2 1513.5 1495.8 2694.7 Pipe Throughput 347.7 349.4 348.7 1059.8 Pipe-based Context Switching 313.7 298.1 318.3 - Process Creation 377.6 377.5 375.7 225.3 Shell Scripts (8 concurrent) 1826.7 2551.7 2333.8 738.3 System Call Overhead 362.2 361.6 363.1 716.2 “cc1.4xlarge”は、計算処理もそうですが、ファイルコピーのパフォーマンスが素晴らしいですね。
あと、やはり、c1.medium(2.5ECU x 2)より、m1.large(2ECU x 2)の方が、CPUパワーが優れていそう(CPUの型番的には納得)なのと、m2.xlargeのI/Oパフォーマンスは”High”ではなかろうか？と思った次第。
Amazon EC2性能検証！気になるパフォーマンスをインスタンスタイプやリージョン毎に計測・比較してみた - RX-7乗りの適当な日々
m1.smallなどの対応は以下の通りです。
その他詳しい計測結果はリンク先をご参照ください。
＊Standard Instances
Small Instance [m1.small] (32-bit) (*Default)
Large Instance [m1.large] (64-bit)
Extra Large Instance [m1.xlarge] (64-bit)
＊High-CPU Instances
High-CPU Medium Instance [c1.medium] (64-bit)
High-CPU Extra Large Instance [c1.xlarge] (64-bit)
＊High-Memory Instances
High-Memory Extra Large Instance [m2.xlarge] (64-bit)
High-Memory Double Extra Large Instance [m2.2xlarge] (64-bit)
High-Memory Quadruple Extra Large Instance [m2.4xlarge] (64-bit)
＊Cluster Compute Instances
Cluster Compute Quadruple Extra Large Instance [cc1.4xlarge] (64-bit)
