LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: TurboLinux cluster

To: Wensong Zhang <wensong@xxxxxxxxxxxx>
Subject: Re: TurboLinux cluster
Cc: "Stephen C. Tweedie" <sct@xxxxxxxxxx>, Joseph Mack <mack.joseph@xxxxxxx>, Alan Robertson <alanr@xxxxxxxxxxxxx>, Cliff Miller <cliff@xxxxxxxxxxxxx>, lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Lars Marowsky-Bree <lmb@xxxxxxxxx>
Date: Thu, 28 Oct 1999 21:34:15 +0200
On 1999-10-29T02:29:06,
   Wensong Zhang <wensong@xxxxxxxxxxxx> said:

> That statements has last for 5 months, however Mike <wanger@xxxxxxxxxx>
> told me that they used the exact patch from LVS project at the end of
> this September.

I just checked, the SRPM includes the 0.5x patch, but doesn't apply it. It
ships ip_cs.[ch] instead.

Well, it is an implementation of the code I would say ;) It appears it can
only do weighted round robin scheduling and doesn't do masquerading, but only
tunneling/direct.

My personal guess would also be that it isn't as scalable as ipvs, but thats
just an uneducated remark after briefly glancing over the code.

If you want to check out just the server code for yourself, I made it
available as ftp://ftp.teuto.net/pub/devel/lmb/cluster-kernel-4.0.4.tgz so you
don't have to grab the entire 17MB srpm.

My comments on this issue are: TurboLinux managed to act in a way which was
perceived as nice, and right now, their code is behind the LVS/ipvs code
featurewise for sure and, as I would assume, not as proven in the real world
nor as scalable.

Having two independant implementations, both GPL'ed, is generally a good
thing, because you can afford one of them to fail and learn from their
errors...

As far as I know, LVS/ipvs is also ahead w/ regard to becoming a netfilter
module, which will be a major boost again for performance and modularity.

Sincerely,
    Lars Marowsky-Brée
        
--
Lars Marowsky-Brée
Network Management

teuto.net Netzdienste GmbH

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