High availability architecture for Apache reverse proxy

Hi,

I'm looking for some advice on ensuring availability of an apache reverse
proxy in an unusual setup.

We host about 50 Tomcat servers in two data centers (more or less randomly
ditributed) where each customer has his individual instance of a web
application. Currently each customer accesses his application using
http://servername/applicationXY (where XY is a sequential number).

Our plan is to introduce a single URL space like http://ourservice/accountname
and also terminate https with a frontend of Apache reverse proxies. A frontend
will proxy requests for servers in the same data center with RewiteMaps (which
are generated from the database of another application), and redirect requests
for the other servers to the frontend in the other data center. The data
centers are geographically differently located and have no direct connection,
therefore we don't want proxying from one data center to the other. The
frontends in both locations will have the same domain.

Note that the Apaches don't do load balancing for the Tomcats, each request
always goes to the Tomcat that contains the specific customer's application.
So I don't need a sticky session mechanism based on session ids in URL or
cookies. Also there is no caching involved, as most of the requests are either
Web Service calls or they are redirected externally to customer's systems.

Now my question:
What do you recommend to ensure the availability of the Apaches within one
data center? I will need at least two Apache machines per data center, but I
don't expect a load higher than one of our usual machines could handle. So in
principle, a hot standby/failover system would suffice. On the other hand, a
load balancing system that detects and ignores a dead node also provides
failover, makes better use of the second system, scales better from start and
is easy to extend to more machines in case the load increases more than we
foresee now. A separate load balancer would need to be redundant as well
though, which means additional hardware costs.

Can you recommend any load balancing software that would run on each
Apache node without need for additional hardware? Or do you have any other
comments/advice regarding this problem?

Thanks
Rainer Frey




------------------------------------------------------------ ---------
The official User-To-User support forum of the Apache HTTP Server Project.
See for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org