Once any one filer resource is nearly consumed, performance will be affected.
My favorite command is "sysstat -u 1", and we see slowdowns when any one or more of the conditions below are occurring continuously (values are approximate; filer is a 3050c):
- CPU: above 90%.
- Net kB/s out: above 75% of the agreggate wire speed of the filer's network connections. (We are read-mostly.)
- Disk kB/s write: more than 10 seconds at a time over 10,000kB/s.
- Cache hit: anything less than 95%.
- Disk util: above 40-50%.
Rohit wrote:
Hi Folks
I had some open questions and i hope you'll help me answer them based on your experience.
- If you just had a single metric to determine filer performance what
would it be ? Would it be CPU usage, throughput, latency, IOPS..or anything else ?
- At your place, what metrics do you use to determine if a filer is
still good enough not to worry about ?