Message boards :
Number crunching :
Question about CPU
Message board moderation
Author | Message |
---|---|
Send message Joined: 16 Oct 07 Posts: 2 Credit: 85 RAC: 0 |
Was wondering, are the works sent by @home sites Beowulf compatible? |
Send message Joined: 2 Sep 04 Posts: 378 Credit: 10,765 RAC: 0 |
Was wondering, are the works sent by @home sites Beowulf compatible? The last time I checked (2004), OpenMosix was able to run Boinc. At that time the boinc client felt that there was one big processor, so my two machines only ran one boinc task. I'm not the LHC Alex. Just a number cruncher like everyone else here. |
Send message Joined: 16 Oct 07 Posts: 2 Credit: 85 RAC: 0 |
Was wondering, are the works sent by @home sites Beowulf compatible? Well yeah, I so want to build myself a Beowulf and then let it crunch the numbers with its terabytes. But I have yet to find any good instructions about doing so... :P |
Send message Joined: 6 Jul 06 Posts: 108 Credit: 661,871 RAC: 196 |
Was wondering, are the works sent by @home sites Beowulf compatible? The Boinc Project team looked at the "Super Host" idea with all other nodes passing data through just one node (the only one talking to the project), but went no further with it. Some people/universities/schools using Boinc have been using a programme called CONDOR that does something like this with software so that only one computer actually does the communicating with the project but I think each node does get its own id from the project as per normal. I have also heard of a user who currently is crunching on QMC@home who uses scrips to make a cluster of computers pass all data through the one node, just check the Top Computer list in QMC and look at the number one computer with an RAC of over 60,000 on a listed 2 core computer. This may be of help. |
Send message Joined: 1 Sep 04 Posts: 137 Credit: 1,691,526 RAC: 383 |
As I recall, the problem with running BOINC on a cluster is that the core client uses shared memory to communicate with the science application and this does not work across nodes on a cluster. So yeah, you would have to have a BOINC instance running on each node, even if you come up with some way of merging them into looking like a single computer to the project(s). - A member of The Knights Who Say NI! My BOINC stats site |
©2024 CERN