This is a follow-up to a comment on the blog:
currently we made sure Riftsaw has no issues being deployed to a JBoss cluster, we are now working to also take advantage of clustering. For load-balancing soap requests we may always suggest using apache httpd, or a hardware loadbalancer but if you're interested discussing this then please start a thread on the forum.
I was meaning that we kinda already have a load balancer (stateless, one can suppose that it's an Apache httpd with mod_jk and no sticky sessions, but may also be a dumb hardware lb) and a jBPM-BPEL based application clustered using JBoss TreeCache 1.4.
I'd like to know whether migrating to RiftSaw 2.0 (since jBPM-BPEL is quite slow and CPU intensive) and maybe Infinispan 4.0 will not cause some mysterious problems resulting from e.g. RiftSaw assigning the same sequential IDs to different process instances or variables that happen to have been instantiated simultaneously on two or more cluster nodes.
Just to let you know, in the 2.0.0.Final release, the riftsaw doesn't have the clustering feature, we had a jira task (https://jira.jboss.org/jira/browse/RIFTSAW-5) for your track. In this release, we only tested that riftsaw can be deployed in the jboss cluster environment, but itself doesn't have load-balance, fail-over features for now.