im not an expert for the Nimble boxes and my english is bad.
But i would like to know what sort of changes you made in December? It looks like the access pattern changes at the end of December. You have an very high percentage of random accesses. That seems to be normal under VDI deployments.
Can you change the Time in the Monitor-Performance view to november - > december so that we can compare the IO-Profile with an view from december -> january?
If there ist mutch more random access, the CPU get more bussy. A change in the kind of data could also hit more cpu because it can be better (easyier) or worser (harder) to compress.
When the compression ratio of these volume is under 10% you could try to build a new volume and deselect the compression feature in the perormance policy.
Thanks for your reply!
I opened a case with support on this issue. We logged in using the 'root' account (only Nimble Employees have access to this P/W)
Running TOP from that session we could monitor the fact that there was consistently 90% or more idle on the CPU
Support and Infosite Technicians are looking into why the graphs are so remarkably out from each other
All of my volumes are showing a minimum of 25% compression - sometimes more depending on the job of the volume.
I'll update this thread once I've heard back from Support on the issue and its resolution.
I'm getting unexpected results from info site based on the IO that I'm doing through the single CS220x2 array.
Here's my story. I've got approximately 300 VMs across 3 hosts that are VDI desktops.
These VDI desktops are primarily Link-Clone based images, with a scripted refresh happening at 1:00AM every day.
This array is used in Higher-Education and sits at about 1000 to 1500 combined IO throughout the day on an average day, with a peak of about 4000 combined IO when the refresh is running (early in the morning)
I've attached a couple of screenshots, one of my Info-site Metrics showing that I'm getting 80% + CPU usage spike when before it was 20% or less
The other attachment is a picture of the metrics off the array over the last 90 days showing the bursty activity, but reduced IO and and Throughput requirements
I expected to see no change in the CPU usage because I reduced the amount of IO and the Throughput (on average). Maybe someone has seen this before - although it's hard to say.
I'm worried because it was hoped that I'd be getting nearly 500 desktops onto this array, but it looks like I may not be able to get too many more than I already have.
Thanks for any replies!
Sr. Systems Administrator II
College of New Caledonia
Prince George, BC, Canada