EC2 推出 ARM 架構的機器...

摘要： 看到AWS 推出使用ARM 架構的EC2 instance 了：「New – EC2 Instances (A1) Powered by Arm-Based AWS Graviton Processors 」。 在 Quick Start 頁面有...

看到ofollow,noindex" target="_blank">AWS 推出使用ARM 架構的EC2 instance 了：「New – EC2 Instances (A1) Powered by Arm-Based AWS Graviton Processors 」。

在 Quick Start 頁面有Ubuntu 18.04 (ARM) 可以選，開起來後操作跟標準的Ubuntu 差不多... 連進去後uname -a 可以看到是 aarch64：

ubuntu@ip-172-30-2-207:~$ uname -a
Linux ip-172-30-2-207 4.15.0-1028-aws #29+nutmeg8-Ubuntu SMP Tue Nov 20 02:59:41 UTC 2018 aarch64 aarch64 aarch64 GNU/Linux

然後來看硬體規格，從最大臺的a1.4xlarge 來看是 16 vCPU + 32 GB RAM，定價是 $0.408/hr (這邊都拿us-east-1 比較)。

對照 General Purpose 的m5.4xlarge 是 16 vCPU + 60 GB RAM，定價是 $0.768/hr。如果看記憶體比較接近的m5.2xlarge 則是 8 vCPU + 31 GB RAM，定價是 $0.384/hr。

對照 Compute Optimized 的c5.4xlarge 是 16 vCPU + 68 GB RAM，定價是 $0.68/hr。

實際跑一些測試，包括 md5、sha256 與 aes (最後 aes 這個通常都有硬體加速)，都用-mutli 16 跑。

ARM 的a1.4xlarge ：

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
md51064103.21k2484789.44k4436178.60k5521370.45k5943380.65k5963896.15k
sha2562059690.93k5652827.82k 11792656.30k 16108863.15k 18086602.36k 18250851.29k

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
aes-128 cbc1593981.48k1723960.38k1752940.20k1767321.94k1770212.01k1768281.43k
aes-192 cbc1400010.40k1496414.83k1516962.99k1527643.82k1529834.15k1527563.26k
aes-256 cbc1222067.79k1296972.50k1313348.18k1321350.83k1322947.93k1321850.20k
blowfish cbc1384982.01k1500548.63k1529793.02k1540091.22k1540937.05k1540767.74k

Intel 的m5.4xlarge ：

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
md51370869.17k3276978.66k5929591.13k7441276.93k8026330.45k8071796.05k
sha256592719.47k1325135.04k2506009.09k3184234.50k3455729.66k3480365.74k

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
aes-128 cbc1060996.96k1121951.87k1135376.21k1141487.96k1143270.06k1143499.43k
aes-192 cbc890438.97k934047.98k943446.44k947576.49k948857.51k949026.82k
aes-256 cbc768686.53k800152.85k806883.93k809804.12k810784.09k810937.00k
blowfish cbc1735490.97k1884059.78k1923876.10k1932711.94k1934477.99k1928680.79k

Intel 的c5.4xlarge ：

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
md51501870.92k3593434.20k6503591.25k8162811.90k8804605.95k8855147.86k
sha256650179.22k1453635.18k2749318.83k3492912.13k3791164.76k3818105.51k

type16 bytes64 bytes256 bytes1024 bytes8192 bytes16384 bytes
aes-128 cbc1163898.98k1230539.07k1244414.63k1252080.98k1254110.55k1254206.12k
aes-192 cbc976610.23k1024570.03k1034886.66k1039442.26k1040872.79k1041143.13k
aes-256 cbc843184.42k877695.30k885125.46k888408.06k889503.74k889766.83k
blowfish cbc1877162.34k2056925.74k2107008.26k2119893.67k2121979.22k2115720.53k

這些數字頗有趣的... 看起來 ARM 上面應該有對某些演演算法加速，使得常見的情境會快很多。不過如果是其他應用的話看起來就會比較辛苦了... 目前就價錢來看未必有絕對的優勢，還是得看應用來決定。