Does anyone understand if the new Xgrid can be used for folding? Or can it? I downloaded it but I only have my 1 G5 here so not much use ( for now ). Do you think Stanford will set something up for Xgrid and folding?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Xgrid and folding?
- Thread starter Selecter
- Start date
- Sort by reaction score
I am new to xGrid but I talked to someone at macworld today who said that the text based client could work. You need to set up a plugin with xgrid and fill in the blanks? I was wondering the same thing. I still dont have a clue of how to do that.
I m not sure it will work.
I mean basically the folding client is running a core, and it has got to have the result of the "frame" before, to be able to calculate the next one(s).
I don't see how xgrid should help. As far as I know, a frame can't be devided. It s the same problem as using a second cpu for the calculation.
If you get it to work however, i d be pretty interested. As you only have to install one client and have it configured, and the other mac s are helping to crunch as soon as they are switched on
I mean basically the folding client is running a core, and it has got to have the result of the "frame" before, to be able to calculate the next one(s).
I don't see how xgrid should help. As far as I know, a frame can't be devided. It s the same problem as using a second cpu for the calculation.
If you get it to work however, i d be pretty interested. As you only have to install one client and have it configured, and the other mac s are helping to crunch as soon as they are switched on
I asked the head guy at folding this and he said that it has been "discussed internally", whatever that means. it would have to have a client set up to recieve a WU and then either break it into chunks for all the CPU's to eat off of or they could set it up so that xgrid could automatically WU to cpus in the grid on at the time. I dont know the best way to handle it, but I do know it's valuable tool that *should* be exploited to good use by folding.
Yes, this can be done on other platforms with endless tinkering and investment but xgrid could be done by any mac user with a few macs around the house or in college labs, etc. So it should be supported by Stanford.
Yes, this can be done on other platforms with endless tinkering and investment but xgrid could be done by any mac user with a few macs around the house or in college labs, etc. So it should be supported by Stanford.
i talked to an apple rep about xgrid yesterday at macworld. it seems like it would have more uses for other non-distributed command line applications. by this i mean folding is already set up to be utilize CPU power from multiple machines in its own way.
As I understand it, the client right now is basically a serial operation program, where each WU computes on each CPU. Xgrid would almost have to work on a parallel approach instead where many CPU's would crunch one WU at the same time, finish the one WU, then upload the result and download the next.
So instead of having each CPU crunch each WU in turn, the whole WU would be crunched by all the CPU's in the grid.
You could do WU's in 10 minutes.
(with enough CPU's)
I sure hope they do it.
So instead of having each CPU crunch each WU in turn, the whole WU would be crunched by all the CPU's in the grid.
You could do WU's in 10 minutes.
(with enough CPU's)
I sure hope they do it.
Okay I kind of asked this in the other XGrid Folding thread but I'll rephrase it here.
What would be the advantage of this over just running the client on each machine.
It seems to me in the case of the XGrid say you have 5 machines all working on one unit and you can finish it in 2 hours.
On the other hand if you have all the machines working you could have 5 work units completed in 10 hours.
The point is in both situations you end up with the same amount of CPU time so you end up with the same results in the long run.
Don't forget that distributed computing is in a way a non localized cluster. What sort of processing power do you think that the F@H project can crank out in 24hours compared to that of Virginia techs super computer? I bet F@H has at least 5,000 times the processing power.
What would be the advantage of this over just running the client on each machine.
It seems to me in the case of the XGrid say you have 5 machines all working on one unit and you can finish it in 2 hours.
On the other hand if you have all the machines working you could have 5 work units completed in 10 hours.
The point is in both situations you end up with the same amount of CPU time so you end up with the same results in the long run.
Don't forget that distributed computing is in a way a non localized cluster. What sort of processing power do you think that the F@H project can crank out in 24hours compared to that of Virginia techs super computer? I bet F@H has at least 5,000 times the processing power.
I just took the time to download the Xgrid preview and read through the documentation.
For the scenario where a single WU would be worked on by multiple machines simultaneously via Xgrid; it will require Stanford to make some adjustments. Hence, someone mentioned that they (Stanford) were discussing it internally.
As it stands now. It can be used for folding farm management, and quite nicely I might add. It could continuously feed full-time folding machines, keeping track of work done in the process. I assume it could also notify you if folding jobs start failing on a node as well. If a part time folding machine comes on-line/available, the controller could feed it a new folding process (or two) automatically.
Regards
For the scenario where a single WU would be worked on by multiple machines simultaneously via Xgrid; it will require Stanford to make some adjustments. Hence, someone mentioned that they (Stanford) were discussing it internally.
As it stands now. It can be used for folding farm management, and quite nicely I might add. It could continuously feed full-time folding machines, keeping track of work done in the process. I assume it could also notify you if folding jobs start failing on a node as well. If a part time folding machine comes on-line/available, the controller could feed it a new folding process (or two) automatically.
Regards
I seriously do hope that Stanford will release the necessary plugin for XGrid. It offer us possibilities for having our slower Mac's helping the Faster Mac's when it comes to folding. I mean, my Powerbook 550 could help the G5. Not by taking half of the work, but a small part of it. This small part of it is defined by the CPU speed, since the G5 would have to wait for the G4 to finish half of the WU (Faster computer/More work). If for example the G5 handled 80% of the WU and the PB handled 20%, it would be perfect, since it wouldnt stress out the PB too much (If the powerbook as faster, then it would take 30%, all of this with the idea of keeping them in sync, so none of them would wait for the other). I seriously dont know if this idea is doable, do any of you guys think something like this would be possible?