Forum Discussion

Walter_Kacynski's avatar
Walter_Kacynski
Icon for Cirrostratus rankCirrostratus
Dec 06, 2016

TCP profile configuration for VIP-targeting-VIP

Here is my setup...

 

Client -> VIP (APM Enabled) -> LTM Policy -> VIP (Application) -> Pool (Members)

 

I am using the default "tcp-mobile-optimized" profile both client and server side connections for EACH virtual server. When deployed into PROD, I noticed the TMM memory increased by 1GB when I went to the VIP-targeting-VIP configuration. When running a single VIP configuration, the memory usage was lower.

 

Are there any recommendations for modifying how the tcp profiles should be configured for the external SS and internal CS tcp configuration? I want to take advantage of the features in tcp-mobile-optimized but I don't want to waste unnecessary memory at the same time. I'm not sure how the Proxy Buffer Low/High values effect the traffic flow when multiple VIPs are involved in the conversation.

 

application delivery
BIG-IP
LTM
profile
tcp
tcp profile
tcp-mobile-optmized
vip

3 Replies

Sort By
  • Vijay_E's avatar
    Vijay_E
    Icon for Cirrus rankCirrus
    Dec 06, 2016

    Use WAN optimized for client side & LAN optimized for server side as a test. This should cover most real-world cases.

     

  • Kevin_Davies_40's avatar
    Kevin_Davies_40
    Icon for Nacreous rankNacreous
    Dec 06, 2016

    If you are doing VIP targeting VIP on the same F5 device then the default TCP profile will be just fine. There is no need for any customised TCP in this instance. If they are on different F5's in the same DC then the lan-optimised profile on the server side of VIP1 as it talks to VIP2 would likely be useful.

     

  • Walter_Kacynski's avatar
    Walter_Kacynski
    Icon for Cirrostratus rankCirrostratus
    Dec 16, 2016

    I did more testing on this and and found considerable performance differences if the server-side connection uses the lan profile as opposed to the mobile profile. This is most noticeable under load when loading a large javascript application. I did not notice a change in memory however so it must have been something else to cause the memory to go up so much.