[Spread-users] Spread for HA rpc services

Colin Meyer cmeyer at helvella.org
Wed Aug 8 13:07:51 EDT 2007


I am considering using Spread to divvy up the load between a number of
"rpc" (a custom protocol, not sun-rpc) servers in a highly available
cluster. The goal is that a client need not be aware of the
configuration of the rpc server pool, but can always easily find a
live server.

The basic idea is that each server would join and listen to the same
queue. The first available server would grab a request message, do the
processing and return the results to the rpc client. Only one server
should respond to any particular request message. 

However, after reading the docs, I get the idea that Spread may not be
the appropriate technology. There does not seem to be a method of
removing a message from the queue before all listeners receive it. Is
this understanding correct?

If Spread doesn't fit the bill, what other technologies should I
consider? Wackamole looks like it would work nicely, but it is also more
complex to configure than a simpler Spread layer.


More information about the Spread-users mailing list