Forum Discussion

Maneesh_72711's avatar
Maneesh_72711
Icon for Cirrostratus rankCirrostratus
Mar 07, 2018

Spinning vCMP Guests

Has anyone observed High CPU concerns when spinning vCMP guests in vertical striped manner in below fashion picking one core from each of the slots ? Post the guest is spun it shows 1 core as Primary/Active and rest of he cores as secondary/standby

 

As compared to choosing all the cores horizontally from the same slot ?

 

When I had spun guest in vertical fashion picking one core from each of the slots for just plain vanilla guest with no config and no traffic passing through it the CPU was constantly on 40% and when I chose it horizontal using 4 cores from same slot it came down to 13%.

 

What are the restrictions and concerns in both the approaches ?

 

2 Replies

  • Bear in mind that you're doing different things here. If you allocate four cores from one slot, you're allocating four cores and four cores worth of memory (the exact amount depends on your hardware to the guest. If you allocate one core per slot you are allocating one core to the guest, and setting up three additional slots to help with the traffic.

     

    You want to ensure you have enough cores per slot allocated that if you lose a blade, the remaining blades can handle your full traffic load. Each slot essentially acts as a full guest handling a portion of the incoming traffic under the direction of the primary slot. So this is completely expected.

     

    This is detailed here: https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/vcmp-administration-viprion-13-1-0/3.htmlguid-f13cc385-46b0-40ac-b695-871d7791a1fe

     

    To reiterate: Cores allocated on a single slot increase the cores and memory available to that guest. Cores allocated on additional slots increase the number of slots available to the guest. Both will increase the ability to handle traffic, but in different ways.