LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Fundamental performance limits of a linux basedloadbalancer/server s

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Fundamental performance limits of a linux basedloadbalancer/server system?
From: Dave Jagoda <dj@xxxxxxxxxxx>
Date: Tue, 11 May 2004 16:12:16 -0700
Joseph Mack wrote:

To rephrase: A load balancer that operates on Layer 7 and maintains
persistent connections to the real servers has to act as a TCP client and is
subject to the port limitations. Is my thinking correct?

I guess so, but do you know that this is how it works?
I think you could have it work that way, but you could also avoid doing so. I believe what Alteon does when doing L7 balancing is to look at the relevant info (e.g. cookie) and then decide which real server to send the traffic to. Once that decision is done, they are just forwarding packets like any router would, and thus aren't actually subject to any concurrent open port limits. Once the decision of which real server to send it is made, it doesn't get changed until the TCP connection closes.

Now if the L7 load balancer does something like aggregate connections (e.g. clients send in connections to LB, then LB keeps an HTTP/1.1 persistent connection open and multiplexes those client requests over that persistent connection), then you would have to worry about port limitations.

The tricky word here I think is 'persistent'. You can get 'stickiness' without having a persistent connection.

Joe
-dj

<Prev in Thread] Current Thread [Next in Thread>