Mar 27, 2026 - For details about updated CVE-2025-53521 (BIG-IP APM vulnerability), refer to K000156741.

Forum Discussion

ukhan20's avatar
ukhan20
Icon for Cirrus rankCirrus
May 06, 2026

CPU/vCPU sizing on rseries

Hi ,

 

I am working on sizing an F5 rSeries platform for a telecom-scale deployment and need expert guidance on CPU/vCPU allocation and platform selection.

 

Traffic profile:

 

L3/L4 traffic (AFM use case): ~100 Gbps

L7 traffic (AWAF use case): ~50 Gbps

 

 

Questions:

 

What is the recommended approach to estimate required vCPU for such mixed workloads?

Is r10900 sufficient, or should we consider multiple appliances or VELOS?

Any best practices for tenant sizing and separation for AFM + AWAF?

 

Appreciate any real-world sizing guidance or reference architectures.

1 Reply

  • Jeff_Granieri's avatar
    Jeff_Granieri
    Icon for Employee rankEmployee
    May 13, 2026

    Hi ukhan20​ 

     

    One approach to consider is to break down the L4/L7 requirements on a per vCPU basis.  Starting off with LTM only this is a raw idea of what you would need vCPU (LTM Only)   

    L4 190Gbps/36vCPU (r10900 max vCPU allocations) = about 5.3Gbps/vCPU  
    L7 190Gbps/36vCPU (r10900 max vCPU allocations) = about 5.3Gbps/vCPU  

     

    We aren't even taking into account AFM or AWAF vCPU requirements.  Applying some base multipliers rough estimation.

    AFM at 100Gbps  L4 100/ 0.6 x 5.3Gbps = 31vCPU approximation

    AWAF at 50Gbps L7 50 /  .25 x 5.3Gbps = 38 vCPU approximation

     

    I don't think a pair of a r10900 would be sufficient.  You might want to look at r12K's and pairs of them as well.  If your looking for a chassis solution I think Velos CX1610 with BX520 blade or CX410 w/ multiple BX110 blades would be best since you can easily upgrade by adding blade capacity without requiring network/topology changes.