I agree that it's a "function" of the servers at the source and destination
side of the ip session, however, the load balancer needs to be configured to
allow DSR (or whatever), in order for it to be able to happen in the first
However, I wouldnt say radwares triangulation was something "entirely
different" - I would say its exact purpose was definately to enable
asymmetric load balancing. It allows servers to bypass the Radware box on
the return trip - connecting instead, directly to the client machine. Thus
creating a triangle... client initiates connection to Radware box, Radware
box passes connection to server, server returns connection directly to the
----- Original Message -----
From: "Alex Samonte" <asamonteIZZATsitesmith.com>
Sent: Friday, November 24, 2000 1:56 AM
Subject: Re: [load balancing] L3 capable load balancers?
> On Fri, Nov 24, 2000 at 12:33:20AM -0000, Will Jones wrote:
> > many L4 switches provide this mechanism - they just all call it
> > different
> > eg.
> > Alteon -- direct-server return
> > Foundry -- direct cache-server return
> > RadWare -- Triangulation
> radware's triangulation is something entirely different.
> It's much more akin to arrowpoints packet encapsulation stuff (or i should
> say the other way around since radware had this functionality first).
> direct server return is more a function of the servers, not the
> load balancers. All you are doing is binding to the VIP on the
> web servers and not sending the return traffic through the LB.
This archive was generated by hypermail 2b29 : Sat Nov 25 2000 - 04:11:10 EST