Corporate Home Open Source Home
Syndicate content
Eucalyptus
3 replies [Last post]
l3golas
Offline
Joined: 03/04/2010

Hi, I've just created a very little cloud (for now just a frontend and two nodes) using Eucalyptus (in particular UEC, that is Ubuntu solution with built-in Eucalyptus). I've executed an instance of an image on my nodes and it seems to work correctly.
Now I would like to test my Eucalyptus parallel features, that is I would like to launch a program, for example matrix multiplication, on the nodes and see if parallelism works (it uses both the nodes). The questions are two:
1) About program structure, the system is clever enough to automatically scale calculations on the two nodes or I have to create two threads (for example the first one working on half the rows and the other one working on the other half rows) and launch one per node?
2) Whatever the answer, how can I do? (steps to follow, commands to launch, etc.)

Thanks in advance,

l3golas

l3golas
Offline
Joined: 03/04/2010
Nobody?

Nobody? I make my question in other and simpler words: I have "built" my cloud, now I want to test a parallel application to see parallelism, how can I do?

mithil
Offline
Joined: 02/07/2010
Hadoop on eucalyptus

You can install Hadoop Map/Reduce framework on the cloud.
Since you can get the machines on demand,you can scale as you want :)
Check this out :
http://www.michael-noll.com/wiki/Running_Hadoop_On_Ubuntu_Linux_(Single-Node_Cluster)/
http://hadoop.apache.org/

Also check out the MPIs for parallelism.

l3golas
Offline
Joined: 03/04/2010
Hadoop

Thank you, I'll give it a try!