Stability of load balancing Options

[jkirchne]

Hi

There has been some comments about the stability of this service offering, we are considering this solution and was wondering if the issues regarding this service have been sorted out?

Thanks

[Damon85]

It might be fine so long as you are not high volume. If you are planning to balance traffic that can't be handled by a single server, forget it -- Total Balance won't work very well for you.

[cprompt]

Doesn't that rather defeat the object of a load balancer if it can't handle loads higher than a single server can manage?

[Damon85]

Yes, it does indeed defeat the purpose. The Planet's load balancing (although not specifically mentioned anywhere, go figure...) is apparently intended only for redundancy, not actual balancing of heavy load.

[Urban]

When we tried to use this service a few months ago we were told The Planet were no longer selling it because of stability issues. We had to opt for a Alteon 180e hardware load balancer instead at $600/month + $650 setup which is quite a difference from the $45/month shared solution we planned to use initially.

[planes]

That's true.
I had everything planned for the $45 solution and I've also been told about that (expensive) solution.

I believe that the (only available) LB solution offered by The Planet is one of its weaknesses. All the other competitors offer rather affordable solutions.

(but I'm pretty happy with the rest of the service!)

[Guspaz]

Don't forget that you can still roll your own load balanced solution.

If you simply need load balancing there are tons of ways to do it. If you need failover, that's trickier, but it is still possible to do by designating a single box as the load balancer box. Stripping it clean and putting the bare minimum on it reduces the risk of failure, and TC boxes have a 100% uptime guaruntee, so certainly you can get pretty good stability.

The other solution is to use super-reliable web hosting as your load balancer, to redirect clients to the proper dedicated SM box. I know there are web hosts that specialize in high-priced ultra-reliable (At least five nines) web hosting with their own insanely redundant systems. All you need is a PHP or perl script to divide up the clients for load balancing and not direct people to boxes that have gone down.

[planes]

I wouldn't really care about failover.

The most important would be to split the traffic I have in the only (web)server I have. Session persistence is crucial though.

[Serhat]

Failover seems somewhat inherent though, because if you're using multiple nodes then you have to take into account that one or more may fail. The chance of a node failing does increase with the number of nodes. One thing you can consider, of course, if to use one main server that redirects certain requests to slave servers (e.g. for images).