Correlizer - new project

[AMDave]

This is a hot-off-the-press new project /EDIT - NOT Alpha (see below) - EDIT/ from Hiedelberg
http://svahesrv2.bioquant.uni-heidelberg.de/correlizer/

They have named it after the client they are using which is unfortunate, because other projects I can think of are already using correlizer clients.
However the client name in the BOINC Manager itself is "BioMedial Genome Correlations 1.00"

56+ MB initial download.

1.25 hour initial estimate.
Actual run time 20 minutes.

clients for win x86 only, (may d/l to a 64-bit OS)
linux x86 and AMD64
Mac OS 10.4 and 64-bit Mac OS 10.5+

no stats export yet

team created
http://svahesrv2.bioquant.uni-heidel...y.php?teamid=4

Now for the investigative part in response to Bok's question there:

Very similar client and purpose to the current test client on AlmereGrid TestGrid (statistical correlation of genomes), I think.

Front page "The application has been developed in The Netherlands at the Erasmus Medical Center in Rotterdam" - that could be where AlmereGrid TestGrid is at?
although I never saw references to Erasmus there or at EDGeS. Intriguing.

found it - post#54 - there is reference to an AlmereGrid pdf document that mentions Erasmus:
http://boinc.almeregrid.nl/forum_thread.php?id=24#50

So, the client and purpose is one and the same, except that AlmereGrid implements the test (AlmereGrid TestGridand) production (AlmereGrid) versions and this one in Germany (correlizer) is the production version only.
But clearly they are coordinated efforts.

So the development and testing has already been done, as Vaughan and myself can attest to on TestGrid, meaning that this project comes into being with a full set of tested apps.
The only thing untested so far is the support staff and infrastructure behind this project, and yet I am well aware that Heidelberg Uni is not renown for getting things wrong.

[vaughan]

I'm in too.

If its like Almere Test Grid then its very easy to get work.

Problem with adding yet another project doing the same work (think DNetc and MooWrapper) is they take CPU cycles from the established projects.

[AMDave]

On my linux boxes I have set up a QEMU/KVM virtual machine and given it all-but-one of the cores.
So the host has a full CPU doing the feeding to the GPU without interference.
Moo! Wrapper is about 2 times faster than if all cores are doing CPU work and the GPU is 'starving'.

I have completed 1 work unit on this project and will get back to it later, but for the moment all 22 of my cores are full of priority work.

[NeoGen]

I'm in too... kinda hefty initial download there, let's see how it goes once the big file gets here.

[Nflight]

I am in Too

[liuqyn]

I can't join the team... so sad...


nevermind, got in by joining through boincstats...