Creating sub-cluster within a Coherence cluster

Question

Hi all,
Does Coherence support creation of 'sub-clusters' within a larger coherence cluster - such that certain caches can be configured to run only on these subclusters, and other caches run on the entire coherence cluster as usual.
E.g., suppose my application consists of 3 websphere clusters (under same cell) - each cluster consists of 2 Websphere server instances. Each Websphere cluster has got a specific functional responsbility (e.g., 1 cluster handles the UI, one handles core processing functionality and the 3rd cluster handles links with external legacy systems). Since the functionality itself is 'partitioned' - its possible that certain data managed by a particular WAS cluster should only be managed within that cluster and not across all 6 WAS instances.
So - in this case - suppose I do have an 'outer' Coherence cluster of all 6 WAS instances (and some Caches are configured to be acessible to all 6 WAS instances - since the data managed in these caches is needed by all 6 WAS instances). Can I configure a smaller Coherence cluster to be available only on say 2 of the Websphere instances (say the WAS cluster which handles legacy links) - and configure certain caches which are available only on this smaller sub-cluster.
regards,
Sanjeev.

Answer 1

I am quite confused about the purpose of the service-name. How would you tie down a cache to a particular service? In the context of the above example, the requirement seems to be:
CacheA should be spread over the UI cluster.
CacheB should be spread over the legacy cluster.
CacheC should be spread over the global cluster.
Are you suggesting something like the following:
<u>Cache config file on a UI node</u>:
<cluster-config>
   <caching-scheme-mapping>
      <cache-mapping>
         <cache-name>CacheA</cache-name>
         <scheme-name>ui</scheme-name>
      </cache-mapping>
      <cache-mapping>
         <cache-name>CacheC</cache-name>
         <scheme-name>global</scheme-name>
      </cache-mapping>
   </caching-scheme-mapping>
   <caching-schemes>
      <distributed-scheme>
          <scheme-name>ui</scheme-name>
          <service-name>ui</service-name>
     </distributed-scheme>
     <distributed-scheme>
          <scheme-name>global</scheme-name>
          <service-name>global</service-name>
     </distributed-scheme>
   </caching-schemes>
</cluster-config><u>Cache config file on a legacy node</u>:
<cluster-config>
   <caching-scheme-mapping>
      <cache-mapping>
         <cache-name>CacheB</cache-name>
         <scheme-name>legacy</scheme-name>
      </cache-mapping>
      <cache-mapping>
         <cache-name>CacheC</cache-name>
         <scheme-name>global</scheme-name>
      </cache-mapping>
   </caching-scheme-mapping>
   <caching-schemes>
      <distributed-scheme>
          <scheme-name>legacy</scheme-name>
          <service-name>legacy</service-name>
     </distributed-scheme>
     <distributed-scheme>
          <scheme-name>global</scheme-name>
          <service-name>global</service-name>
     </distributed-scheme>
   </caching-schemes>
</cluster-config>The basic question seems to be: how do you control the nodes over which a cache is spread, purely from the cache name?
Also, the 3.2 <role-name> feature seems to be something that addresses this requirement. How does that play v/s a service-name?
My requirement is similar (needing to control the nodes over which different caches are spread) but I do not quite understand how the service-name would be used to satisfy this example. Could you please explain via cache configurations for this example?
Thanks
Ghanshyam