Hello,
I wanted to provide an update to are ongoing process. Over the weekend and during this week we began adjusting socket and core assignments. Add or remove and then run some load sim software to generate and then of course having people use the linked clones has been great. It seems that 2 sockets and three cores each has been improving performance or at least moved the bottleneck someplace else. As you can see by the below graph we are still seeing a lot of yellow.
Currently there are are 12 dedicated systems up and running with 8 in use, 62 linked clones with 38 in use. There is good news here at least we are starting to get above the 50% mark which we never were before, and some of those small green slivers are 4 to 8 VMs running in 25 to 80 MHz.. One by one I powered off our blades, disabled hyperthreading, rebooted, re-enabled hyperthreading, rebooted, and then brought the host back into the cluster. Probably didn't do anything but it was a thought. We also have a case open with NetApp. We are running 60%+ CPU usage with sub1.5K IOPS on our FAS series with a flash pool and no one can seem to find out why. We did find some cabling issues that will be resolved over the weekend as well.
We are still scratching our heads as to why so much yell when clearly there is plenty of CPU.
Jim
