At 08:40 19-12-02 -0500, Carlos Ordonez wrote:

You can use LVS (Linux Virtual Servers) as a load balancer and high
availability clustering technique.
Didn't LVS use a heartbeat to check the presence of the worker servers?
If so, then I think that is not good on z/VM unless you have techniques
to shutdown and autolog servers on demand so that you never have servers
that would otherwise be idle.

I admit that I have not looked in enough detail yet at all knobs and
tweaks, but I believe most of such load balancing schemes tend to
round-robin over
the workers. That would scale less well than a stack or push-down list
(like VM TCP/IP uses with multiple servers on a port) because it keeps all
penguins around rather than try with as little as you can.

Rob

Reply via email to