On 10 Jul 2012, at 14:28, Vladimir Blagojevic wrote:

Mircea,

First of all thanks for the feedback! Your ideas and pointers are
appreciated as always!

On 12-07-10 7:02 AM, Galder Zamarreņo wrote:

Before blogging, it would it be good do some testing to measure the performance boost with some sampled data?

The advantage of doing this, apart from proving that the boost is indeed present, you can show the code pattern that users need to use when storing a bunch of key/values which is proven to work.

I agree with this approach Galder. I entertained this idea as we need to
move large amounts of data during intermediate phase in map/reduce
efficiently without causing lock timeouts and deadlocks. I'll get back
to this later as I am currently working on cacheloader for map/reduce
https://issues.jboss.org/browse/ISPN-2037
+1. Thanks Vladimir for looking into this. This problem was raised by our SAs yesterday in the ISPN's training in Berlin and people had a lot of interest in it. I've raised a JIRA so that we can track it: https://issues.jboss.org/browse/ISPN-2156