Terracotta Clustering for solving the "continuation-serialisation-problem" in a cluster environment?

[Gabriel Gruber <Ga...@workflow.at>]

Hello dear Experts, 

We are currently implementing our cocoon 2.1.10/spring/hibernate based 
application at a big customer in a clustered environment on tomcat 5.028 
and JDK 1.4x using Apache 2.0 and mod_jk as load-balancer. As recommended 
we use "sticky-sessions" due to the fact, that serialisation of 
session-state (and their nested flowscript continuations) is not possible. 


However there seems to be a new star on the sky of clustering named 
terracotta which virtually builds a "grid of jvms". Terracotta has finally 
gone open source and offers their product on the base of TPL which is a 
slightly modification of the Mozilla Public Licence (MPL): 
http://www.terracotta.org/confluence/display/orgsite/Licensing 

As a specific usecase http session clustering is well documented in the 
terracota documentation. In contrast to "classic" http clustering 
solutions, terracotta does NOT serialize the http session-objects but 
rather "shares a tree of objects between cluster nodes". 
http://www.terracotta.org/confluence/display/docs1/Sessions+Quick+Start 

Might this finally be a solution for having a seamless failover for cocoon 
users with flow continuations, cforms models and all that stuff? 

Gabriel 

______________________
Mag. Gabriel Gruber
Senior Consultant
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Workflow EDV GmbH, Dannebergplatz 6/23, A-1030 Wien