1 Reply Latest reply on Sep 23, 2010 1:38 PM by Galder Zamarreño

    moving from JBoss Cache 3 to Infinispan 4.1

    Daniel Cullender Newbie

      Hi,

       

      We are attempting to migrate from jboss cache 3 to infinispan 4.1 and have a couple of questions we would like to ask around replicating the Fqn grouping.

       

      We have seen that infinispan provides a TreeCache which replicates this structure but we have read there is a performance impact occurred when using this API. Are there any performance gains from moving over from jboss cache to the infinispan TreeCache structure? If so, are they significant gains? Furthermore, Infinispan TreeCache does not seem to notify listeners when data is put into the cache - Has anyone else got this to work?

       

      To avoid using the TreeCache API and use the Cache APi directly, we have to change our entire cache structure. Is any way to replicate the FQN grouping using the Cache API directly? We have tried a couple of things (for example: using Cache(String,Map<Object,Object>) causes write skew) but with limited success.

       

      Thanks

        • 1. Re: moving from JBoss Cache 3 to Infinispan 4.1
          Galder Zamarreño Master

          Daniel Cullender wrote:

           

          Hi,

           

          We are attempting to migrate from jboss cache 3 to infinispan 4.1 and have a couple of questions we would like to ask around replicating the Fqn grouping.

           

          We have seen that infinispan provides a TreeCache which replicates this structure but we have read there is a performance impact occurred when using this API. Are there any performance gains from moving over from jboss cache to the infinispan TreeCache structure? If so, are they significant gains? Furthermore, Infinispan TreeCache does not seem to notify listeners when data is put into the cache - Has anyone else got this to workΩ

          There're huge performance improvements in Infinispan compared to JBoss Cache, regardless of TreeCache itself. There're improvements as well in the way Infinispan implements TreeCache structure. Instead of keeping hashmaps of hashmaps, the entire tree structure is kept in a single map which is a lot more memory efficient.

          To avoid using the TreeCache API and use the Cache APi directly, we have to change our entire cache structure. Is any way to replicate the FQN grouping using the Cache API directly? We have tried a couple of things (for example: using Cache(String,Map<Object,Object>) causes write skew) but with limited success.

          By FQN grouping I suppose you talk about keeping a map of things mapped to a key. If this is so, Cache.put(key, Map) should do the trick. What's the write skew issue you had? Could you post a unit test?