microsoft / microsoft/pai

Enrich cluster GPU utilization report

Open
#5,403 0 comments 0 reactions 1 assignee View on GitHub

@suiguoxin is already working on this.

Since Mar 31, 2021.

pai-dev
Dominant language
JavaScript
Stars
2.7k
Forks
554
Avg merge
20h 42m
Merged PRs (30d)
14

Description

Cluster Utilization in One Week

Cluster Level​
GPU*Days Used GPU*Days Provided GPU*Days Capbiity Average Number of GPU Cards Provided Max Number of GPU Cards GPU Card Occupied Rate GPU Utilization
Overall 1000 1200 1500 150 200 70% 50%
VC0 (V100) 200 250 250 25 20 80% 60%
VC1 (K80) 800 1250 1250 100 100 60% 40%
User Level​
Number of Users Number of Active Users
Overall 50 15
VC0 (V100) 20 10
VC1 (K80) 40 10

Top 10 Users (Ordered by Resources Used)

User name Resources Used (GPU*hour) GPU Utilization Number of Submitted Jobs​ Number of Runned Jobs​
user0 3000 40% 5 10
user1 2000 50% 20 50
Job Level​
Number of Runned Jobs Number of Submitted Jobs Retried Number / Rate Succeeded Number / Rate Failed Number / Rate Stopped Number / Rate Failure Exit Code Average Waiting Time Max Waiting Time Average Running Time Max Running Time Long-Running Rate (> 48 hours) Short-Running Rate (< 30 minutes)
Overall 300 200 60 / 20% 60 / 20% 60 / 20% 180 / 60% -210(10), 220(5), 404(3) 54 minutes 80 minutes 35 minutes 5 hours 30% 30%
VC0 (V100) 100 100 20 / 20% 20 / 20% 20 / 20% 60 / 60% -210(5), 220(5), 404(3) 2 hours 15 minutes 80 minutes 3 hours 3 minutes 5 hours 70% 10%
VC1 (K80) 200 100 40 / 20% 40 / 20% 40 / 20% 120 / 60% -210(5) 3 minutes 5 minutes 3 minutes 30 minutes 10% 70%

Notes:

  • Retried / succeeded / failed / stopped rates are calculated based on completed jobs
  • Succeed rate + failed rate + stopped rate = 1
  • Waiting Time of Running Jobs = Current Time - job submitted time

Top 10 Jobs (Ordered by Resources Used)

Job name Resources Used (GPU*hour) GPU Number VC Job Duration Job Start Time​ Job Status
job0 3000 40 VC1 (K80) 1 day 10.7 hours 21-03-26 03:14:19 RUNNING
job1 2000 12 VC0 (V100) 23.5 hours 21-03-27 03:14:19 SUCCEEDED

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.