CAELinux in the Cloud

Hi,

it looks like the problem is related to your login details see the line:
" NX> 202 Authenticating user: nx"

you should connect using login = ubuntu and password = caelinux...

Are you sure that you use these settings in your NX session?

Ok, I see!!

You should keep the default Key that comes with NX !
I have tried jsut now and it worked perfectly. So just create a new session without altering the key.

The EC2 key that you get is to login to ssh only and is not mandatory as you can directly login with SSH using username/password;
The EC2 SSH key does not work for NX.

Hi

I have done some benchmark, but it needs to be polished a bit before publishing.

My comparison basis is a pretty fast quad core Phenom II X4 965 at 3.4Ghz with 8Gb RAM so when compared to Opteron 2.4Ghz CPUs used in some instances of Amazon EC2, the frequency difference is quite big.

However, I found that the ratio of single CPU performance between EC2 and local machine just follows the frequency ratio of the CPUs.

So on an EC2 instance with 2.5 EC2 per core (1ECU ~ 1Ghz), you can guess that you are running on a ~2.4Ghz Opteron and thus estimate the "reference" time for a single CPU job.

When you add more CPUs, the parallel speed-up enters in the game and this factor strongly depends on the model size and model decomposition method.
At the moment, I did only small tests so the parallel scaling was quite limited and thus the data is not representative..
I will repeat the procedure with a much finer mesh to see what is the actual parallel performance of the system and post the results here.