VM貸出サービス
VM貸出サービスのデフォルトスペック
マルチユーザで利用できるPodman、Singularityを導入する
Podman
Singularity
コンテナベースの学習環境を提供できるようになる
そこでPodmanをwrappしたie-podmanを作成した
ie-podman run --ip --gpu [IMAGE_NAME]
$ ie-podman run --rm --ip --name test -dit ubuntu:20.04 Assign an IP address: 10.100.20.1 6842a8a572a73f7a21542dbb9fec5b5b4ca668aa9c2f7d7fe2522e3be316836d $ ping 10.100.20.1 PING 10.100.20.1 (10.100.20.1) 56(84) bytes of data. 64 bytes from 10.100.20.1: icmp_seq=2 ttl=63 time=0.215 ms 64 bytes from 10.100.20.1: icmp_seq=3 ttl=63 time=0.201 ms
ie-podman sif [IMAGE_NAME]
$ ie-podman images REPOSITORY TAG IMAGE ID CREATED SIZE localhost/e175733_test latest cb42448d39ca 21 seconds ago 138 MB $ ie-podman sif e175733_test INFO: Starting build... Getting image source signatures (省略) INFO: Creating SIF file... INFO: Build complete: e175733_test_latest.sif $ ls e175733_test_latest.sif
singularity run --nv [SIF_NAME]
sbatch [BATCH_FILE]
$ sbatch test.sbatch Submitted batch job 1318 $ squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 1318 steins test e175733 R 0:01 1 amane
#!/bin/bash #SBATCH --job-name sample #SBATCH --output logs/%x-%j.log #SBATCH --error logs/%x-%j.err #SBATCH --nodes 1 #SBATCH --cpus-per-task 8 #SBATCH --gpus tesla:1 #SBATCH --time 01:00 singularity exec --nv [SIF_NAME] [COMMANDS]
実験環境
実験概要
FROM ubuntu:20.04 RUN apt-get update && \ apt-get upgrade -y
syscall
dd
書き込み速度の比較