is a same WU being run several times?

[De Berghut]

Hi guys,

I was fooling around a little with new clients etc, trying to upgrade my system a little... In the end i ended up two clients working independently, but receiving exactly the same WU, or at least it seemed: same project, same clone, gen and run.. So i ended up with the same question i'm already having for quite a while: i'm i correct thinking an exact same WU gets to go 'out' to be run on several computers? Which would, mathematically be a 'waste' in the ideal circumstances.. But of course, EUE's do happen and all... But still, i wonder..
thanks,
steven

[Rattledagger]

The exact same wu, in other words Project, Run, Clone & Gen, should only be issued to another client in case the 1st. has reported it as an error, or has passed the preferred deadline.

In case a particular client has somehow "lost" a wu, it will be re-assigned the same wu. So, if you're incorrectly used the same host-id on two clients, and you've stopped client-1 before starting client-2, it's possible you've be re-assigned the same wu. If so, chances are you've only be credited once...

[toTOW]

This might also occur in some rare occasions when two clients request a WU simultaneously.

[codysluder]

This might also occur in some rare occasions when two clients request a WU simultaneously.

I doubt that's true. Do you have a source for that informaiont?

It can happen if a WU passes the preferred deadline (causing a duplicate to be issued) and then the first WU has an EUE, causing the partial results to be uploaded (and causing the first copy to be reissued).

[De Berghut]

Well, when i started the 2nd client, the first one was still running, on this particular WU, which it had downloaded, let's say 14 hours before.. The 2nd client downloaded a WU in the meanwhile, which appeared to be the same one... Anyway, there were no issues such as EUE's, nor any deadline passed or anything.. But i was just wondering if the server, sends out the same WU several times, just to make sure it at least gets one WU back in time and in 'one piece' as soon as possible... But i guess i'll be satisfied with your answer: "The exact same wu, in other words Project, Run, Clone & Gen, should only be issued to another client in case the 1st. has reported it as an error, or has passed the preferred deadline."
But now i think about it, do i have to give each client a different machine ID?

[bruce]

Yes, each client on a single system must have a different UserID. As far is it was concerned, the server reassigned the same WU to the SAME machine because it requested a new WU without finishing the one that was assigned to it. (Because of the duplicated MachineID, the server knew it was the same client. You're the only one that though they were independent clients.)

[De Berghut]

Thanks, one always learns new things

[Tigerbiten]

I've just got 2 GPU work-units with the same Project ( Run, Clone, Gen) numbers.
Machine ID 3

Code: Select all

[23:36:20] - Digital signature verified
[23:36:20] 
[23:36:20] Project: 5506 (Run 3, Clone 197, Gen 186)
[23:36:20] 
[23:36:20] Assembly optimizations on if available.
[23:36:20] Entering M.D.
[23:36:27] Working on p5506_supervillin_e1
[23:36:27] Client config found, loading data.
[23:36:27] Starting GUI Server
[23:37:41] Completed 1%
[23:38:54] Completed 2%
[23:40:07] Completed 3%

Machine ID 6.

Code: Select all

[23:36:19] - Digital signature verified
[23:36:19] 
[23:36:19] Project: 5506 (Run 3, Clone 197, Gen 186)
[23:36:19] 
[23:36:19] Assembly optimizations on if available.
[23:36:19] Entering M.D.
[23:36:26] Working on p5506_supervillin_e1
[23:36:26] Client config found, loading data.
[23:36:26] Starting GUI Server
[23:37:40] Completed 1%
[23:38:53] Completed 2%
[23:40:06] Completed 3%

I hope I get full points for crunching the work-unit twice.

Luck ...........

[De Berghut]

So, may i conclude there actually is a 'hole' in the system? It sure would be quite a waste if a lot of computers are doing the exact same calculations
Or is it possible you changed the machine ID's just before?

[toTOW]

Or it's a bad WU, and will fail twice (I don't know, there's no data in the stat DB yet).

[Tigerbiten]

So, may i conclude there actually is a 'hole' in the system? It sure would be quite a waste if a lot of computers are doing the exact same calculations
Or is it possible you changed the machine ID's just before?

They've both run around 800 work-units each.

Code: Select all

Launch directory: C:\FaH\Data3
Arguments: -verbosity 9 -gpu 2 

[21:29:08] - Ask before connecting: No
[21:29:08] - User name: Tigerbiten (Team 33)
[21:29:08] - User ID: 347F33C6101F890E
[21:29:08] - Machine ID: 3
[21:29:08] 

Code: Select all

Launch directory: C:\FaH\Data6
Arguments: -verbosity 9 -gpu 5 

[21:29:14] - Ask before connecting: No
[21:29:14] - User name: Tigerbiten (Team 33)
[21:29:14] - User ID: 347F33C6101F890E
[21:29:14] - Machine ID: 6

Its the first time I've seen the same P(R,C,G) at the same time.
Before I've seen it due to a bad work unit, where another box got the same work-unit after an EUE.

Luck ...........

[De Berghut]

I'll just have to wait untill i can work with Nehalem (i7) at an affordable price till i can use my gpu i've got a good graphics card, but this good old cpu doesn't have SSE2 yet i guess fah isn't interested in making gpu-clients work with SSE1? *sigh*

[Tigerbiten]

I dont think the GPU client needs the SSE2 instruction set.
Which card do you have and which vid drivers do you have loaded.
If you've the correct vidio drivers loaded give it a try.

Luck .............

[De Berghut]

I've got a hotfix driver, just can't remember which version, but i do recall it was the most recent one let's say 3-4 monhts ago...
About the SSE2 instruction set: i started a thread about it a few months ago, and i got a reply from one of the AMD-guys: viewtopic.php?f=42&t=3569#p33937
So sadly, i'm afraid i'll just have to suffer waiting a little longer :'(

[De Berghut]

Well guys, i'm happy to tell you my GPU is running now, on my athlon xp 3200+ And the gpu activity (slightly overclocked even) is pretty much about 99% constantly I can sleep again now :p