[cisco-voip] UCS memory sizing best practices

[Ryan Huff]

Here is Cisco’s virtualization software guide which is sorta like the bible of all the do’s and dont’s for Cisco virtual servers with VMWare: https://www.cisco.com/c/dam/en/us/td/docs/voice_ip_comm/uc_system/virtualization/virtualization-software-requirements.html

I’m not sure if the specific statement you’re referencing exists, but what you will find in that guide is the requirement that no Cisco virtual sever is subject to resource over subscription on the virtual host server.

Applied to your scenario; if I have 8 Cisco UC VMs using 8GB of RAM each on a single host, I’m required to have at least 64GB of RAM available at all times on the host server, just for those VMs (not including RAM needed for the host itself). The same rule applies for CPU and storage (i.e; even though thin provisioning may work, you should always thick provision).

At 80+% utilization I’d suspect you may be right at the line if that doesn’t include the host resource overhead. I’d do a resource review on the host and triple check nothing is oversubscribed.

If you have a vCenter and the enterprise licensing setup, the resource oversubscription conversation usually includes it black sheep cousin, DRS.

Together, the two shuffle VMs around as needed, in real time, to “smooth out” the resources usage with in the host cluster / vCenter. With oversubscribed host resources, DRS tends to be more “active”. The dynamic resource scheduler is a HUGE no no with Cisco UC VMs; they cannot move at all, while powered on.

Hope this helps,

-Ryan-

On May 4, 2018, at 17:00, Daniel Clark <daniel at scopsblog.com<mailto:daniel at scopsblog.com>> wrote:

Hello folks,

We have a customer with a C220 M4S running ESXi 6.0.  It's currently sitting at 42GB of 48GB memory utilization (87%).  We had to increase the RAM allocated to multiple VMs when we upgraded the customer's UC environment from 11.0 to 11.5 and 11.6.  This drove it over the 80% warning threshold configured in our monitoring system.  This sensor is purely looking at memory allocated versus total memory on the physical host.

The utilization has been as solid as you'd expect since the upgrade, and there are no plans to increase utilization further, but management has asked that we verify Cisco's best practices before we raise the monitoring threshold to 90% to clear the warning.

Does anyone have links to Cisco (or VMWare) documentation which says anything along the lines of, "Memory utilization on an ESXi host should not exceed x%"?  We can open a TAC case for this if needed, but I wanted to check if anyone had a quick reference document that my google-fu couldn't come up with.

Thank you in advance!
Daniel
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/cisco-voip/attachments/20180505/9548b17d/attachment.html>