1 Reply Latest reply on Feb 22, 2007 11:47 AM by Manik Surtani

    Merge replication?

    Basil Achermann Newbie

      Hello,

      We are currently considering using JBossCache in one of our projects. It should be able to do the following:

      --
      There are several producer nodes. Each node collects data that it keeps in memory. Producers read, update, write their own data. They don't need to know (and shouldn't) about other nodes and their data.

      Then there are consumers. They need to see all data combined for processing and/or visualization. Lookups must be possible. Consumers must be able to sync immediately after connection loss or crash.
      --

      So what's the best way to go about this?

      Is it possible to create one big cache and configure regions that are shared by a subset of members only?

      Is it possible to set up several caches, one for each producer, which are merged into a single cache or something that looks and behaves like one on consumer nodes? Consumers shouldn't have to worry about multiple cache instances and about where to look for specific data.

      What kind of communication overhead is to be expected if we were simply using one cache among all producer and consumer nodes? (10-20 producers, 2-5 consumers)

      Thanks in advance!

        • 1. Re: Merge replication?
          Manik Surtani Master

           


          Is it possible to create one big cache and configure regions that are shared by a subset of members only?


          Why? Shouldn't consumers see everything?

          Perhaps what you need is each producer uses a STANDALONE cache, using a TCPCacheLoader.

          The consumers create a cache cluster which replicates everything. The consumer-cluster also has a TCPCacheServer which the TCPCacheLoaders write to.