3PAR StoreServ Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

IOPS information needed, how to get it

Hazem KUNNANA_1
Occasional Visitor

IOPS information needed, how to get it

Hi, I was asked to get the following values from our 3par.


• Average Write Throughput (IOPS) –
• Average Read Throughput (IOPS) –
• Peak Write Throughput (IOPS) –
• Peak Read Throughput (IOPS) –
• Average Write IO size (KB) –
• Average Read IO size (KB) –
• Sequential Write (%) –
• What time does the Peak start? –
• How long does the Peak last (HH:MM)?

 

I have the "HPE 3PAR StoreServ Management Console 3.0" installed and connected to the system.

Now the question is how can I get this information from that software. I know I look as a lazy person, but i hope you excuse my lack of experience and the fact that this info is required ASAP.

2 REPLIES
Dennis Handly
Acclaimed Contributor

Re: IOPS information needed, how to get it

Have you looked at the CLI commands statvlun/histvlun or srstatvlun/srhistvlun?

They won't tell you about when peaks start and end.

HPEStorageGuy
HPE Blogger

Re: IOPS information needed, how to get it

Unfortunately, there is no metric telling which percentage is sequential vs random I/O, but everything else you can get from 3PAR metrics.

 

The first option that comes to mind is to launch the “Performance Analysis” from service processor for the busiest period you want to analyze.  Or 24h if you want to get an idea about one day. Note that it is limited to 1000 samples, so with 1s sampling you get about 17 minutes, and 90 second sampling interval will give bit more than full 24h.  Of course, such long interval will not report peaks, but average over the period, thus 1s is a good one if you want to zoom in some particular period of interest.

 

And… if you want for example 1 second sampling for full 24 hours, that can be obtained from CLI interface, just requires a minor script or logging output to file.  There are scripts to do that.

 

Additionally, there is the on-node system reporter which has all the historical data for the array, at daily, hourly and 5 minute (high resolution) sampling. That is an easy way to take a look back what has been going on – but limited to 5 minute average level of accuracy.  It also has the histograms for various objects, including host ports which tells you exactly the IO load the hosts have been presenting.

 

SSMC can also be configured to graph it, however not sure if you can set an interval less than 5 seconds with it, but maybe that is enough.

 

Here's a link to a System Reporter technical white paper.