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

danieloliveira's avatar
danieloliveira
Icon for Altostratus rankAltostratus
Jan 30, 2023
Solved

vCMP logical interfaces throughput

Hello, we currently have 2 BIG-IP 15800 each one connected with 2 100Gb interfaces. 

So i have  a guest vcmp with 8vCPU and 8 logical interfaces 0.1, 0.2, 0.3 and so on to 0.8. In the cli-console or at my zabbix those interfaces are detected as 10Gb each, and i can see traffic in all of them...

My question is, are those virtual interfaces capped at 10Gb ? Or in another words, how much bandwidth do i have on this vCMP?

 

 

  • The hypervisor is virtualizing networking so you will see an 'interface' for each vCPU.

    For standard VS's , typically a TMM can't process more then 10G of traffic, for fastl4 use cases the FPGA offload occurs before TMM.  

    Said another way, there isn't a hard cap on bandwidth to the guest but there are practical limits in place for most traffic use cases. 

    https://support.f5.com/csp/article/K03740927

     

     

6 Replies

  • The hypervisor is virtualizing networking so you will see an 'interface' for each vCPU.

    For standard VS's , typically a TMM can't process more then 10G of traffic, for fastl4 use cases the FPGA offload occurs before TMM.  

    Said another way, there isn't a hard cap on bandwidth to the guest but there are practical limits in place for most traffic use cases. 

    https://support.f5.com/csp/article/K03740927

     

     

    • danieloliveira's avatar
      danieloliveira
      Icon for Altostratus rankAltostratus

      Thak you Brian_Van_Lieu for the answer and article suggestion.

      "Said another way, there isn't a hard cap on bandwidth to the guest but there are practical limits in place for most traffic use cases. " In essence are the limits of processing per vcmp and the type of traffic/layer going trough rigth ?

      "typically a TMM can't process more then 10G of traffic"  Can i undertand that a single vcmp even with 8vCPUs cant process more that 10G of traffic?

      • Brian_Van_Lieu's avatar
        Brian_Van_Lieu
        Icon for Employee rankEmployee

        TMM is threaded on the cores, so its really 10G/s (or more) per TMM. The 10G/s per TMM is a round-ish number for guidance, not an absolute. 

        If you have 4 TMMs, you can achieve 40Gb/s (or more). 

        This is very dependent of course on traffic, packet sizes, and what additional (if any) work BIG-IP is doing on the traffic as it passes through the data plane. 

         

         

         

  • Hey danieloliveira  - I saw that nobody had answered your question yet, so I featured your post in this week's Highlights article, and will ask a colleague for some help.