|
Ricardo Kleemann wrote:
> Question... are there plans for other balancing algorithms?
new algorythms appear mainly in response to user's demands. Both
the -sh and -dh algorythms were user contributed in the past year.
> So say realserver1 has a small number of connections but something goes
> wrong and it's under high load. It would be nice for that server to either
> no longer get connections, or get a smaller percentage, because it's not
> able to really handle requests due to a high load condition.
The current algorythms have the convenient (from the point of view
of coding) feature that they don't require any information from
the real-servers. The director keeps track of connections etc
and does something based on the connection table. This seems to
have done well enough and load seems to average out reasonably well.
(most clients are making similar requests it seems, or else the
requests require only a small part of the server's resources).
To do what you want requires an agent on the real-server giving
information about load the server (as mon does to give an up|down
for each service). The use of such agents to monitor the servers
was a matter of much discussion on this list in the early days of LVS
(look for "agents" and "load informed"). The final outcome was
that no-one thought the current balanced algorythms bad enough
to stop all their other activities to write one that did
load-informed load balancing.
But if someone writes one I'm sure people would use it.
I don't think it would be that hard to do if you were interested.
Joe
--
Joseph Mack PhD, Senior Systems Engineer, Lockheed Martin
contractor to the National Environmental Supercomputer Center,
mailto:mack.joseph@xxxxxxx ph# 919-541-0007, RTP, NC, USA
|
