Hi All,

I have two Red Hat entreprise Linux 5 servers running the same applications, my goal is to load balance the incoming traffic between these two servers without adding an additional hardware as load balancer, the incoming traffic is basically Web traffic and on the servers, Apache and PHP are running.

Red Hat solutions that I found are:

1- LVS
2- Piranha
3- Red Hat cluster suite.

Can LVS run on one of the real servers (or both servers, for redundancy issues) and load balance the traffic between its server and the other one? If yes, will LVS has a big impact on the CPU/memory usage? How can I estimate this usage if I know how much traffic is coming and the HW specs?

Same questions for Piranha.

Can I install Red Hat cluster suite on these two servers, and configure it in a way that the virtual IP is on different server for different request (taking in consideration that the same session will be always served by the same server)?


Finally, is there any other/better solution that can fit my requirements to load balance the traffic without additional hardware?

Thanks

The first thing you really need to do is assess exactly where your bottlenecks lie. You need to establish performance baselines for all aspects of your system. Lest you find a "solution" to something that isn't even a problem ... and lest you *fail* to find a good solution to *genuine* issues. Your first step is absolutely a good baseline.

Your *second* step is to decide whether the primary goal is "performance" (best overall throughput), "capacity" (if general performance is OK, but you need a solution that addresses peak loads), or "redundancy" (your main goal is to make sure you stay on-line if one or more software and/or hardware components fail).

A great deal of homework needs to be done before you can even *begin* to consider "solutions". Your post gives very little idea of how much of this homework you've done.

There's tons of material on the web, many fine products ... and, yes, many fine open-source solutions. Starting, for example, with the ubiquitous "squid". Once you decide where you need to go ... you've got lots of options for how to get there.

'Hope that helps .. PSM

Hi Paul,

Thanks for your reply! In fact I've checked "squid", it's a cache server more than a load balancer, I'll give more info about my applications and my concern:

1- Applications are based on PHP pages that query an Oracle DB and return results (No need to worry about the data of the DB, it will be replicated on both servers.) Cache servers will not be a good idea for me since data in the DB changes frequently.
2- I'm concerned about load balancing the HTTP traffic when both servers are alive, but also I have to keep the virtual IP alive if one server goes down; since I have two servers and same applications on both of them, I will care about redundancy also.


Hope this will clarify my query and my homeworks

Still any clue about the applications listed previously (LVS, Piranha, RH Cluster suite) will be appreciated.