| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Distributed computing with BOINC Agent |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 7
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'd like to create a distributed computing system, so that all my computers will work on the same job. Is it possible? I've got a LAN and all my PCs can connect to Internet.
I've tried using the host-client system implemented in BOINC Agent (in the menu: Advanced->Select computer), but it's only a remote control system (or am I wrong?). Thanks for the answers and sorry if this topic is alredy listed here. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello pietrop ,
I'd like to create a distributed computing system, so that all my computers will work on the same job. The projects that we are currently running are not set up to do this. The overall task is split up into small work units, each to be run on a stand-alone thread. If you put BOINC on X computers, you can simultaneously run X * Y work units, where Y is the number of cores per computer. So, 1 job per core is the way that our projects are designed to run under BOINC. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello pietrop , I'd like to create a distributed computing system, so that all my computers will work on the same job. The projects that we are currently running are not set up to do this. The overall task is split up into small work units, each to be run on a stand-alone thread. If you put BOINC on X computers, you can simultaneously run X * Y work units, where Y is the number of cores per computer. So, 1 job per core is the way that our projects are designed to run under BOINC. Lawrence Many thanks, Lawrence. Implementing this feature will increase the amount of job done by home users like me. In my case, I've got a macbook pro (core 2 duo) and a PC (pentium 4 HT, significatively more slow in computations), and it's more difficult to terminate a job with the PC, but it would more productive fusing the power of the P4 with the power of the Core2Duo! (actually with the mac it takes about 6-7 hours to terminate a job). Thanks a lot again =) |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
How would it be more productive? The same work is done either way.
World Community Grid don't have exceptionally tight deadlines, so members can complete work at their own pace. You are correct that "Select computer" is for remote administration of BOINC clients. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
How would it be more productive? The same work is done either way. Many computers working on the same jobs will decrease the time spent to complete it, and this is usefull if you can't leave the computers on for hours and hours. This is valid when the CPUs have very different computing capacity. It would be like a parallel working method in a parallel working system!  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No. The existing limitation is the frequency with which applications can checkpoint.
As an administrative simplification, something similar to what you propose has been suggested many times before. It has, however, never been suggested for performance reasons. It is the responsibility of the project technicians to split up the work into suitable sized chunks so that they will run even on computers that are only used a couple of hours a day. An argument can be made for smaller work units. Equally an argument can be made for bigger ones. Practically speaking, World Community Grid have had great success with their target of about 10 hours on an average computer per work unit. Sorry. Your idea won't work, and isn't needed anyway. Don't let this negative result stop you from suggesting any other ideas you may have. Your idea was on the right track, it just had a few faulty assumptions. Right now, the developers are looking at ways to process in parallel on multiprocessors, where shared memory can give a real boost. |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
hi pietrop!
----------------------------------------If you have slow/part time computers that have problems to finish some jobs in time then the Rice project is ideal for them. It is running the same CPU time for every computer, whatever their speed. Obviously the granted credits are not the same, they are proportional to the work that the machine was able to do during that time, so it's fair. Currently the fixed CPU time is still 8 hours for a few days, but starting soon with new batches of work it will be increased to 9 hours. Any computer that will be active enough for providing 9 hours of its CPU time over 12 days (the current deadline for this project) will be able to contribute, even the slowest one. Last, this project takes checkpoints very often, therefore work lost between two sessions is kept to a very low level. I think that crunching on this Rice project is the simplest and most effective way to keep a good old beloved computer busy if you really want so (don't forget the energy cost vs work done ratio  ).And as usual (in case you prefer other projects) the work that it will do for Rice will be less Rice work to be distributed by the servers to multi-project machines, so your preferred project will finally take its share through WCG's balanced distribution. Cheers. Jean. |
||
|
|
|