For more information regarding the security incident at F5, the actions we are taking to address it, and our ongoing efforts to protect our customers, click here.

Forum Discussion

Romani_2788's avatar
Romani_2788
Historic F5 Account
May 17, 2018

You might not have a one size fits all, but some general guidance are indeed necessary.

 

K14760: High density vCMP, does state that for single-core vCPU guests, these only supports:

 

  • LTM standalone* GTM standalone
  • LTM and GTM combination

Combining multiple modules on one vCMP guests requires that it has enough processing power, and enough memory to go round all the provisioned modules.

 

When considering processing power however, it is important put in mind the effect of HTSplit (see K23505424: Overview of the HTSplit feature), meaning what ever number of CPU you allocate to the guest, half will be used by TMM while the others will be used by Control Plane daemons and processes. This is discussed in K15003: Data and control plane tasks use separate logical cores when the BIG-IP system CPU uses Intel Hyper-Threading Technology.

 

So when a vCMP guest is provisioned with 4 vCPUs, 2 will be allocated to TMM, while the other 2 for all other processes.

 

Now provision that guest with LTM/ASM/APM, and you can begin to see how depending on the amount of traffic being pushed through the guests, and what needs to be done in traffic handling, this can begin to reach some thresholds on the guest. And note that such considerations will also have to be made for an appliance as well.

 

It is unfortunately not as easy as sizing it in precise configurations, as a lot of variables would have to taken into consideration. While one implementation might get away with this configuration, with a few virtual servers, ASM policies and moderate traffic, another implementation a higher amount of virtuals, policies and a traffic throughput to match, might not.

 

So, it is no wonder that increasing a vCMP guest from 4 vCPUs to 8, would definitely help to relieve the load if already strained.

 

A guest with 2 vCPUs running LTM/ASM/APM, will be strained from the very beginning, as all ASM and APM processes combined with all the other Host processes will have to share that single CPU, so as a general guidance you want to have at least 2 vCPUs for TMM and at least 8Gb of memory when running LTM with other modules.