Re: [jboss-dev] Renaming server/web to server/jbossweb-standalone
On 02/11/2010 12:45 PM, Dimitris Andreadis wrote:
On a different thread I came to similar conclusions. We have
a) the APIs/capabilities/services (what)
b) their configuration or QoS characteristics (in what context)
For (a) we could have
'web' - ee web profile
'full' - ee full profile
'osgi' - an empty mc/osgi kernel
'tomchat' - standalone jbossweb
For (b) modes like
a) development, single-node,on-demand, unsecured, non-optimized
b) development-ha, multi-node, on-demand, unsecured, non-optimized
c) production, single-node,eager-loading, secured, optimized
d) production-lb, multi-node, eager-loading, secured, optimized,
load-balanced
e) production-ha, multi-node, eager-loading, secured, optimized,
load-balanced, replicated
What's your thinking behind the difference between d) and e) ? That d)
would include the non-HA variant of mod_cluster but none of the other
clustering capabilities?
For Java clients, the load balancing behavior of cluster-aware proxies
can be independent of the HA aspect, but a JGroups channel is still
required. The distributed caches used for replication only start if a
bean is actually configured for replication, and if that's what you want
it's possible to annotate an EJB3 SFSB such that it gets a cluster-aware
proxy but a non-replicated cache. So for remote EJB3 use cases I don't
know how much real world demand there would be for a d) variant.
--
Brian Stansberry
Lead, AS Clustering
JBoss by Red Hat