Hi Daniel This is my view of where we are.
Service lookup - We are moving to a domain model now that doesn't rely on service lookup internally. If you take a look at the workspace code and sample/calculator-distributed as it now stands the service endpoints in the domain are calculated in the workspace, i.e. before the individual composites are deployed and don't need looking up by each of the nodes that are running composites. It seemed like a simpler solution as the service lookup piece was causing unwanted complexity. The workspace code is new so there aren't docs yet. There is the calculator-distributed sample though and the Tutorial code gives it a good workout. Ask here if you want to know more :-) Load balancing and failover - The default plan for load balancing is to run composites in a cluster. I got close to having a simple sample running on my own machine in a Tomcat cluster with the old domain approach. I haven't converted this over to the new workspace yet but here is how I expect it to work. - The details of the head of the cluster are provided to the workspace as a node to which composites can be deployed - This configuration causes references targeting services that will run in the cluster to have their bindings configured with the URI of the head node of the cluster - All of the real nodes in the cluster request the same composite from the workspace and hence start the same services but of course on different physical addresses - Service requests are dispatched to the head node of the cluster based on the configuration provided by the workspace - The head node sprays out requests to the cluster as appropriate. This should (!) work for stateless service but I haven't actually tried it yet. Stateful services, e.g. conversations or services with COMPOSITE scope are a little more problematic as the problem of affinity and fail over comes up. To make fail over work in the stateful case we will have to provide persistance of the conversation state etc. This is still not there yet Hope that helps and please feel free to jump in if you want to help out with any of this. Regards Simon On Tue, Apr 8, 2008 at 3:19 PM, <[EMAIL PROTECTED]> wrote: > Hi all, > > I just read through some older postings concerning Load-Balancing, > Failover and Service lookup. > > http://www.mail-archive.com/tuscany-user@ws.apache.org/msg01858.html > http://www.mail-archive.com/tuscany-user@ws.apache.org/msg01539.html > > I'm interested in the status of these topics ? > What are the current ideas/plans or are there already any solutions ? > What about the Simon's toughts > http://cwiki.apache.org/TUSCANYWIKI/distributed-runtime.html > > Bye, > Daniel > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >
