Point of entry - always the weakest link?

Question

I've been reading/researching load balancing for a project and some guides talk about having a load balancer and a backup load balancer which use heartbeat to check on each other. When one goes down the other takes over etc...

However, if the amount of traffic is enough to take down your first load balancer how much of a chance does your second one have? It doesn't make sense to have a second beefier machine waiting around in case the first one goes down. I'm sure this is fairly naive question, I'm guessing that the amount of work the load balancer does is so trivial that you will be able to see an upswing of traffic and be able to take care of that as necessary etc..

Also, is using a backup load balancer pretty typical?

I'm using the latest version of Ubuntu with Apache 2 on the rackspace cloud for my servers.

Answer 1

The secondary load-balancer is not there for when the primary can't handle the load, it's there as a redundant fail-over should something happen to the primary which would make it inoperable such as needing maintenance that will take it offline, a failed power supply, failed hard disks, burned out components, and so on.

Having redundant hardware in general is pretty common practice for just about anything that could disrupt a system from a high-level device to a low-level component of those devices: redundant domain controllers, load-balancers, web servers, database servers, firewalls, routers, switches - then there are redundant hard disks, power supplies, network cards, fans, and other components that could act as a single point of failure for a any given system.

In terms of secondary "backup" load-balancers, yes, this is quite typical because your web requests will hit the load-balancer first, then get distributed out to a web server. If a load-balancer goes down - those requests don't get distributed.

This could be useful to read:

How can I balance incoming web traffic amongst N apache servers?