powered by Jive Software

Clustering: HA or Load Balancing

Looking at the roadmap there’‘s an Enterprise version ‘‘future’’ entry to support clustering. I see there’'s a related issue (ENT-43) currently targeted for 3.4 and a bit of recent SVN activity.

Looking at the threads, people seem to be wanting the ‘‘clustering’’ feature to solve one of two problems:

  • High Availability: transparent fail-over to a hot standby server.

  • Load Balancing: horizontal scaling to support more users in a single domain.

Clearly there is a lot of overlap on these two things: load balanced clusters often - but not always - support transparent fail-over.

So my question is: What scenario is the clustering work targeted at solving (first ;-)? HA? Load balancing? Both?



p.s. My first post to the forums, so I feel I ought to say what a great piece of software Openfire is and that the forum post-editing thing is a real pleasure to use!

Welcome to the club, Chris!

The first stage of the Openfire Clustering project primarily works towards a load balancing solution. The primary goal is to support a larger user base.

As you said, solving one issue will help you in the other one as well, so you can expect improvement in availability as well: if one of the nodes fail, for example, others will remain available, which obviously improves availability. HA hasn’'t been fine tuned yet, in that extend that client sessions are migrated off to another node right before a node is being shut down, for example.

This first stage isn’‘t the end of the road, clustering wise. Improvement of and additions to functionality will be added in future stages, but I’'ll leave it up to the actual devs to discuss that.

Thanks for the response Guus. Much appreciated.

As a dev guy myself, I never like answering these kinds of questions, but: do you have a rough idea of when the clustering work will be released?


We’‘re hoping to have a first beta release out in about a month. We’‘re currently estimating that a production release will be ready by early Fall. We’'ve kept a lid on all the progress made so far, but will start posting up a lot more details soon.