Redundant web servers.

Dennis Muhlestein devel at muhlesteins.com
Mon Sep 17 09:08:50 MDT 2007


>
> Is purchasing more hardware an option? If so, buy more RAM and continue
> running what you have now. It sounds like it's working, but how much more
> uptime are you looking for?

Not initially, I've got to get something in the next couple weeks that
is "Faster" and "won't die under load".

>
> But, if you want to rip and replace (or reconfigure) then my recommendation
> is to do the least complicated: Invest in a "proper" load balancer, run the
> web servers on physical machines, and sync or async data with clustering
> software to a beefed up box that is running a VM of each physical box
> (many-to-one).

I'm thinking of doing a hybrid at this point.  I'm going to take our
services that don't need as much memory/processing power.. things like
blogs or small informational sites and consolidate them to a couple
servers.  I may replace Xen with OpenVZ so I can host more services
per machine.  Right now, the bottleneck really is memory.

Then I'll take the other machines and run the services without virtual
machines.  I can cluster them as the ibm article demonstrated. I like
your idea of running a master machine with virtual servers and then
syncing those to the physical machines.  I'll have to look into that
as I move stuff around and see if we have enough hardware to do
something like that.

>
> I can't directly help with the load balancer (I have a distributor that
> can), but can with the server hardware if the need comes up and you're not
> against running Supermicro.

I'll keep that in mind.

Thanks for your input!

-Dennis



More information about the PLUG mailing list