POJO Mojo

Check out this thread on EclipseZone: http://www.eclipsezone.com/eclipse/forums/m92134186.html

I am not sure about r-OSGi or what Eugene is thinking of in terms of integration, but he is met with a touch of resistance quoting the always-astute Martin Fowler's edict around avoiding object distribution as well as the research at Sun.

There is an extremely important point to be made here:
1. Most people misunderstand the paper's assertions. I intend to talk to him about this next week in London. I want to see if I understand more accurately what the intent was.
2. Corba != DSO. Nor is Corba or RMI or EJB in any way related to DSO.

I will have to write an article on the differences. Don't get me wrong. Fowler's paper is correct. Fear of abstractions around remoting is well placed. But Terracotta is more like NFS--I like to call it network attached memory--than it is like Corba. And, NFS has the ability to safely abstract file I/O. Why then can we not abstract heap I/O? Doesn't Virtual Memory exist in the OS?

More later.

I recently cohosted a webinar with Filip Hanik who is PMC lead on Tomcat and Tribes for Apache. (You will need the Raindance player to view this.)

The gist is that Terracotta recently ran ourselves through JBoss's own benchmark for a customer and found that while JBoss was doing 10 - 100 ops per second (depending on the load script used), Terracotta was doing 2000 - 4000. It is worth looking at the webinar to learn about the use case, the customer's architecture under JBoss and their desired architecture goals. The webinar does go into detail on how Terracotta works and perhaps why the performance differences.

I am very excited to have helped this customer. The benchmark differences translated to the real-world use case. The customer now needs less application code (many tricks were being done to work around JBossCache's performance) and fewer servers in order to handle the same volume of web user traffic.

I am going to write an article shortly on how Terracotta is delivering such high transaction volumes. I expect to post it here by end of the week.

--Ari

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.

Read on for my summary...