TCP profile – client and server

Originally posted on 22-Feb-2008 byDavid Sherman

I have read through this forum topic with great interest. I found several messages that dealt with tcp profile optimization. I have one question though.

When configuring a TCP profile, there is an option to select Protocol Profile (Client) and Protocol Profile (Server). The default is TCP as the client profile and (Use Client Profile) as the server profile.

If configuring a VIP that will front end a pool of high end web servers that will be accessed from the Internet, should the Protocol Profile (Client) be set to the tcp-wan-optimized profile and the Protocol Profile (Server) be set to tcp-lan-optimized profile? At least as a start?

0Rate this QuestionAnswers to this QuestionUSER ACCEPTED ANSWER&F5 ACCEPTED ANSWERUpdated 22-Feb-2008Originally posted on 22-Feb-2008 byJason RahmF5I would think so intuitively, and this is the direction I received a few years ago from F5 support. However, in recent months, our ops team has received direction that the client profile in a WAN environment should be carried through to the server. Best advice is to test,test,test.0USER ACCEPTED ANSWER&F5 ACCEPTED ANSWERUpdated 22-Feb-2008Originally posted on 22-Feb-2008 byMarcus SlawikF5hi dsherman,

i have made the best experience with putting the lan-optimized profile on the Server part and the wan-optimized on the client side. so i have a lot of stuff local in my data center where bigip and servers are close together whereas my clients are far away also via lines with higher latency and it really speeded things up for the users in comparison also with users connecting directly to the servers.

schwiddy0USER ACCEPTED ANSWER&F5 ACCEPTED ANSWERUpdated 25-Feb-2008Originally posted on 25-Feb-2008 byDavid Sherman0Thanks for the info. It is really helpful. I wish though, that the F5 developers would step up and give some guidance on this. The TCP Express capabilities are prominently advertised, however, I find that configuration recommendations are lacking. It would be very helpful if the developers would say, If your web servers are high performance (Im using IBM P590), and your clients all connect on high speed LAN connections, then generally best performance would be the LAN protocol to both client and to server. If the clients are on slower links or on the internet, then generally, best performance would be with WAN optimized on client and LAN optimized on server.

Perhaps the recommendation would be to use L4 and utilize the PVA and forgo F5 SSL termination would be best, if the servers are up to the task.0USER ACCEPTED ANSWER&F5 ACCEPTED ANSWERUpdated 04-Mar-2008Originally posted on 04-Mar-2008 byJames ThomsonF5The performance of your web server doesnt matter necessarily. The conditions of the LAN are what matter. So, if you are on a local segment, use the TCP-Lan-optimized for the server-side and the WAN-optimized for the client-side.

If you have a high performance server in Hong Kong and your BIG-IP in London and youre accessing it over an internet link, then youd want to use the wan-optimized in that case.

Using FastL4 and not taking advantage of the dual tcp stacks just means that your High Powered Server is going to have to use the least common TCP options that each client supports individually and will need to negotiate different tcp options for each client. So, if some clients are coming in with windows 2000 on a dial-up connection, then your server will spend resources lowering window sizes, dropping packets and being inefficient.

With the LTM full proxy in line, your server will always get optimal tcp options from a tcp-wan-optimized profile on the server-side of the BIG-IP LTM and the clients each get handled as best they can with the wan-optimized.0You must be logged in to answer. You can loginhere.Specify an image to upload:Choose ImageCloseInsert ImagePost NotificationYour post has been identified as spam. If this is not the case, please co.CloseAbout DevCentral

We are a community of 250,000+ technical peers who solve problems together.