POJO Mojo

I spent last week with Floyd and crew at QCon 2007 in London. The conference was highly energized and lots of brilliant people were hanging out discussion Java and real application design. It was very refreshing to not talk as much about "the future" and whether or not Java is dead and about new languages in the runtime. There was a strong return to use cases and pragmatic advice for the architect and lead developer.

Our little table was busy with people who either attended one of my talks or those who heard about us during Rod Johnson's Spring talk. It was really interesting to see all the big company names (mostly European companies at this London-based conference) on people's badges. It was more interesting to see how much these people were struggling with what they thought was the state-of-the-art in open source clustering (very few were using proprietary clustering products).
"My peer-to-peer solution doesn't work at all. It splits the brain regularly."

"My peer-to-peer orphans nodes all the time."

"I built my cluster by hand and it just does not scale. I don't want to maintain it anymore."

In one of my talks, I reviewed a recent customer experience with JBossCache and what they found in migrating to Terracotta. You can view an (hour-long, sorry) webinar with Filip Hanik, PMC Lead on Tomcat and Tribes, and me discussing this exact same topic. Since this talk was in the smallish Solutions track, we had 35 attendees. They seemed enthusiastic about migrating from various session replication schemes and other clustering projects onto Terracotta as soon as this week. In fact, 2 revealed that they were already users of our open source project and had only needed our forums to get up and running on their own.

My second talk was the usual "intro to Terracotta" and this was attended by upwards of 80 conference attendees. This walk was FUN. I had other speakers in the talk (now that we are open source, they want to know what it is...very kewl). And lots of amazing questions.

"What happens with interned strings?"

"What about volatile with respect to your batching and windowing?"

All smart questions. All smart people. I walked away feeling very challenged and very much engaged with the group.

I also found Peter Kriens in attendance at my talk. He and I agreed to find a way to more tightly couple OSGi and Terracotta as quickly as possible. Wouldn't it be neat to be able to drop Terracotta back out of an [already instrumented] application without shutting down the cluster? It is pie-in-the-sky I recognize, since disabling clustering means EVERY cluster member has been orphaned and you can't rejoin a cluster once that occurs but still interesting nonetheless.

As a result of our presence at QCon, we saw an extra 100 downloads in a single day (not bad for 400 attendees). 10 large companies asked for help in beginning Proofs of Concept where they are looking to replace brittle and / or other OSS clustering with ours.

Next steps for Terracotta? Lots more conferences. See our homepage at Terracotta.org for more details. Also, it appears time to start talking to the Java community about helping lessen the complexity of the application stack. Specifically, how things like clustered web frameworks (Wicket, Rife, JSF, Struts 1 & 2) without serialization and with object identity will help lighten our collective development workload.

--Ari