Performance of each machine in the Cluster

Question

Lachlan Simpson 2015-6-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/225043-performance-of-each-machine-in-the-cluster

评论： Lachlan Simpson 2015-6-26

I am looking for some performance metrics for my cluster. Unlike this question I am only looking for very simple measures to report on:

 - how much uptime of each node (worker+head)
 - CPU/Worker usage ("how much quiet time? Are we saturated?")
 - average queue length ("how busy is the cluster?")
 - average number of CPUs/job ("what type of jobs?")
 - average memory usage/node (worker+head) ("do we have enough, or do we need more?")
 - all of these over time

This list isn't exhaustive, but as you can see it's relatively simple stuff. Each of these things can be measured per machine, but I would presume that MatLab/MDCS has some sort of built in measures or plugins that will centrally log this info for mining?

Just to be 100% clear: I am not interested in the performance of any particular clustering algorithm, any particular job that's run, any particular measure of how far apart the CPUs are in a job or anything like that. Literally only "how hard are my servers working, how utilised are my workers and worker nodes, across time."

The across time is important.