[Beowulf] JVM Clustering

C. Bergström cbergstrom at netsyncro.com
Wed Feb 20 13:01:32 PST 2008


Hi all..

First post to the list so lets hope I get this right...

With the recent discussion about the relatively low cost of SDR IB it
made me wonder if I could solve these two very broad goals

1) Distributed jvm to run very large reports/statistical analysis
2) load balancing/hpc type solution for the app container

Based on my initial searching online I've found jessica2 [1] which seems
to solve the most difficult parts, but certainly open to
ideas/suggestions on better approaches.  I'm also somewhat familiar with
tc (terracotta) which will mostly do what I need. (When active-active
support comes out later this year.)  I do feel much more comfortable
merging the C code of jessica2 against the latest version of kaffe a lot
more than trying to debug java bytecode..  Something like jessica2 also
has long term application maintenance benefits over tc.

----
I'm also interested to hear what others are doing for inexpensive/power
efficient nodes that have pci-e, but this is probably a future
discussion.
----

I just want to send a very big and warm thanks to all those that make
this list so great!


/Christopher


[1] http://i.cs.hku.hk/~clwang/projects/JESSICA2.html




More information about the Beowulf mailing list