Is there shortage of longer WU's?

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0
Topic 191361

Over the last week at least i had little louck recieving enough work to keep the computer busy and mostly its because (you can look and see for yourslef) I've had almost no longer WU's. My results pages are awash with units that have struggled to take 300seconds. Maybe should go back to standard app. Don't want to put extra strain on the system.

I'll soon be back on seti so that'l solve the problem of not having enough work but is there a problem resulting in there not being many longer WU's?

Don't mean to sound like moaning. i think my machine is sucking up its daily quota and spitting it out faster than new work can be sent. I suppose i should sit and wait whilst 'S4' project is worked throught then we'l see what happens in 'S5'

Michael Karlinsky
Michael Karlinsky
Joined: 22 Jan 05
Posts: 888
Credit: 23502182
RAC: 0

Is there shortage of longer WU's?

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

RE: Bad luck? Yea

Message 37365 in response to message 37364

Quote:
Bad luck?

Yea probly.

Been thinking that last couple of days. seriously if you look theres hundreds of them.

Maybe its a sighn were approaching the last legs of 'S4'. Doing all the little bit that were missed. :)

Probly just bad luck.

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

OOooo. Just got a batch that

Message 37366 in response to message 37365

OOooo. Just got a batch that should take about 10-15min each. 2 at same time so about an hours worth of work in total. i'm Increasing BBC-CCE priority.

Ocean Archer
Ocean Archer
Joined: 18 Jan 05
Posts: 92
Credit: 368644
RAC: 0

That's one nice thing about

That's one nice thing about BOINC-based systems, if one project is struggling, there are always others that are willing to use your spare computing power....


If I've lived this long - I gotta be that old!

JoeB
JoeB
Joined: 24 Feb 05
Posts: 124
Credit: 90502700
RAC: 30309

Problem fixed?? I had a

Problem fixed??

I had a computer with a bunch of 20 min wus interspersed with a few 60 min wus. I ran out of work earlier this evening so I detached and reattached. The work that was then downloaded up was 6 hours per wu! So plenty of work for that machine!

I think the short wus were hurting the project also in that it generated a lot of server workload. IMHO it is more efficient to have fewer & larger wus, and my perception is that the project has achieved that. I'm on dial-up so it was a double benefit for me.

I think your problem is in the process of being solved by the project.

Joe B

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3562358667
RAC: 0

Each workunit consists of the

Each workunit consists of the same amount of observational data. The length difference is because at low frequencies there is less potential detail extractable about the angle the candiate signal's coming from. This results in less needed crunching. To 'fix' the 'short workunit problem' the way you want would require doing one of three things: Do alot of pointless crunching on short work units. Make the workunits contain 4x as much data. Split the short wu's into 2 parts instead of 8. Going back to the first (the old einstien app) is obviously a nonstarter. The second would result in a 4x spread in download sizes instead of crunching times, and might not be possible without significant changes in the science app and WU generators. Doing it would iether require increasing the timespan of each WU 4x, and with s4 only using the best 7% of the acumulated data (600h out of 1 year) finding chunks of good data 4x as long probably isn't possible. The other option would be to look at 4 frequency bands instead of just one. This would require 4 start/stops in the app for each band checked. Spliting each download into 2 wus instead of 8 would probably require fewer changes than making the download files larger, but would still have been more work than just adding a check for the frequency band of the WU and changing the size of a single constant in the science app as a result.

We've been promised a better scheduler for s5, that should give the short wu's to the slowest machines and leave the big ones for faster boxes. Looking at user bandwidth levels for this would also be a nice feature to add.

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

I don't mind much in my case

I don't mind much in my case that there so short but i was atarting to worry maybe something was wrong.

Little worried about this too;

Quote:
max # of error/total/success results 20, 20, 20

Don't tell me thats normal. Just looking randomly at some of the short units i've done-about to do and spotted that in all those i looked at, quite a few. Bit excessive isn't it? or is there something i'm missing there?

Pav Lucistnik
Pav Lucistnik
Joined: 7 Mar 06
Posts: 136
Credit: 853388
RAC: 0

I, for one, would love to

I, for one, would love to have only short workunits. Where can I sign up?

Michael Karlinsky
Michael Karlinsky
Joined: 22 Jan 05
Posts: 888
Credit: 23502182
RAC: 0

RE: Don't tell me thats

Message 37372 in response to message 37370

Quote:

Don't tell me thats normal. Just looking randomly at some of the short units i've done-about to do and spotted that in all those i looked at, quite a few. Bit excessive isn't it? or is there something i'm missing there?

That is normal :)

Michael

Pepperammi
Pepperammi
Joined: 20 Feb 05
Posts: 131
Credit: 437943
RAC: 0

RE: RE: Don't tell me

Message 37373 in response to message 37372

Quote:
Quote:

Don't tell me thats normal. Just looking randomly at some of the short units i've done-about to do and spotted that in all those i looked at, quite a few. Bit excessive isn't it? or is there something i'm missing there?

That is normal :)

Michael

Quote:


in "old" type of WU was
max # of error = 20
total = 20
success results = 20

in new type of WU this parametrs was changed:
max # of error = 5
total = 10
success results = 20

eventually found this post in a link from the link you gave (thanks for pointing me there). These i've got must be "old type units". Sorry i never noticed the figures are that high before, surprised me. Still does.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.