|
Mike Whitaker wrote:
>
> I'm considering deploying LVS on our web clusters (in my not so copious
> free time!), and would appreciate some thoughts on the following:
>
> A typical web cluster for us would look something like this:
>
> LVSdirector--------------+---------------real server----------------\
> | 192.168.0.98 | 192.168.0.99 212.187.214.99 |
> 212.187.214.97| | |
> | +---------------real server---------------\|
> | | 192.168.0.100 212.187.214.100 ||
> | | ||
> | +---------------real server--------------\||
> | | 192.168.0.101 212.187.214.101 |||
> | | |||
> | +---------------real server-------------\|||
> | 192.168.0.102 212.187.214.102||||
> | ||||
> \--------------------------------------------------------\||||
> |||||
> switch
> |
> /
> Internet_____________router_______/
>
> Director using DR, of course. I THINK I have this setup right?
hmm we're reaching new heights of ascii art here.
if 212.187.214.97 is the VIP, then the real-servers will need this IP too
(and not x.x.x.99-102)
> Now, our problem is that we have several of these clusters, which are
> geographically widely separated. We're doing some measure of DNS-based
> providing of the 'nearest' cluster for a given user, but it isn't necessarily
> true that a given cluster can handle all its potential traffic.
there is a geographically based load balancer by one of the LVS developers
which uses BGP
http://supersparrow.sourceforge.net/
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
|
