08-19-2011 09:42 AM - edited 03-01-2019 10:02 AM
I have vSphere 4.x environment that has have IBM 3650's and half B200M2 (One cluster being made up of the 3650's and one cluster being made up of the B200M2's). The IBM machines are 2 socket quad core with 34GB's of memory and the Cisco are 2 socket 6 core with 48GB's of memory. The workloads on the IBM and Cisco ESXi hosts are the same. Mostly Windows 2003 and 2008 machines. On the IBM machines I get around 1GB's to 2GB's of shared common memory (Shared Common memory - Amount of machine memory that is shared by all powered-on virtual machines and vSphere services on host) on each ESXi host. On the Cisco machines I get around 15MB's on each ESXi host. Considering that is one of the great features of VMware (the sharing of memory) why is there such a huge disparity between the machines? Is it because the Cisco blades are NUMA machines? Is there a setting I'm missing? That is a huge difference. I don't regret getting the UCS stuff as I like it but this is big feature of VMware that is not working with the UCS.
Solved! Go to Solution.
08-19-2011 02:48 PM
Please have a look at the following:
Transparent Page Sharing on the Nehalems will not kick in until memory utilization nears 100%.
Edit: should read: Transparent Page Sharing on the Nehalem and later generations will not kick in until memory utilization approaches 100%.
08-19-2011 02:48 PM
Please have a look at the following:
Transparent Page Sharing on the Nehalems will not kick in until memory utilization nears 100%.
Edit: should read: Transparent Page Sharing on the Nehalem and later generations will not kick in until memory utilization approaches 100%.
08-19-2011 03:38 PM
Thanks a lot. That's what I was looking for.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide