Very short WUs


Advanced search

Message boards : Number crunching : Very short WUs

Sort
Author Message
Profile Saenger
Volunteer tester
Avatar

Joined: Sep 13 06
Posts: 125
ID: 79
Credit: 411,959
RAC: 0
Message 4591 - Posted 7 Nov 2008 16:00:05 UTC
Last modified: 7 Nov 2008 16:00:49 UTC

The last couple of days the WUs were very short, I needed several hundred to get my daily ratio even for my quite diverse machine. They used to be about an hour long, later just half an hour, now it's down to a minute for most of them.
I think it's as well some considerably load on the server to serve so many WUs per hour if they only last for such a short period.
Can't you make them considerably longer, by the facto 20 at least, imho better factor 100?

Not something to do before your congress, but soon afterwards would be fine.
____________
Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Profile David Ball
Forum moderator
Volunteer tester
Avatar

Joined: Sep 18 06
Posts: 274
ID: 115
Credit: 1,634,401
RAC: 0
Message 4593 - Posted 7 Nov 2008 17:19:06 UTC - in response to Message ID 4591 .

Hi Saenger,

It's a problem with a particular series of WU going out of bounds on energy and ending the run early. If I understand correctly, it's not actually an error on the client side but something to do with the WU input file being fed to Charmm and the characteristics of the specific protein-ligand for the WU. They're working on solving the problem. See More Here:

http://docking.cis.udel.edu/forum_thread.php?id=366&nowrap=true#4585

I hope I didn't mangle the explanation too much. You really should read what the project people say in the other thread linked above.


____________
The views expressed are my own.
Facts are subject to memory error :-)
Have you read a good science fiction novel lately?

Bob and Claudia Weidman

Joined: Nov 24 08
Posts: 4
ID: 4029
Credit: 211,752
RAC: 0
Message 4674 - Posted 6 Jan 2009 23:04:42 UTC

I'm having the same problem now, a couple months after the initial post in this thread. I will get a WU that shows it should take over an hour to run and then it is complete within 10 minutes......not all of the time, but probably 90% of the WUs are doing this. I'm unsure as to whether or not I am getting any credit since my graph hasn't budged from being a dot on the graph.
I have 5 other projects crunching with no problem at all (with the exception of superlinkattechnion, which hasn't been sending me any WUs.)
Any ideas?
Thank you in advance for your time.

Bob & Claudia Weidman

Rob

Joined: Jan 2 09
Posts: 1
ID: 5461
Credit: 60,999
RAC: 0
Message 4676 - Posted 7 Jan 2009 7:28:39 UTC

Hi---I am new to this project and have been wondering about the very short WUs also. Most are "energy exceeded" errors? Am I actually doing any usable work for the project or are these WUs no good? I DO appreciate the credit as another project I won't name will error out up to 12 hours in and give 0 credit.

Profile [B^S] Acmefrog
Volunteer tester
Avatar

Joined: Nov 14 06
Posts: 45
ID: 252
Credit: 1,604,407
RAC: 0
Message 4677 - Posted 8 Jan 2009 3:08:01 UTC

I will just say that I have had both very short WUs (like 10 min) and ones that take like 8 hours to do (seems very long).
____________

Profile Trilce Estrada
Forum moderator
Project administrator
Project developer
Project tester

Joined: Sep 19 06
Posts: 189
ID: 119
Credit: 1,217,236
RAC: 0
Message 4679 - Posted 8 Jan 2009 21:06:56 UTC

Hi All,

Yes, we have been distributing workunits with short time. Unfortunately we cannot do anything to solve the problem. The short time is due to energy violations that are more frequent in these ligands. The work done is still useful because it gets to conformations that are valid. We analyze then those results that may not be complete but definitively are useful.

You can go and see the ligands we are running at the list of targets and compare the trypsin ligands with the hvi ligands. They are very different in terms of flexibility and that is partially the reason for the energy violations. Right now we are distributing 1k1m and we are only 8 ligands away from the end of this batch. Now I'm shortening the sampling per ligand, then we expect to finish these short workunits in 1 week and a half or 2 max. After that we will run alphaP38 which are more stable, and longer workunits

I'm so sorry for the inconvenience, and we appreciate that you are still with us. The pain will end soon


Buffalo Bill
Volunteer tester

Joined: Dec 19 06
Posts: 6
ID: 338
Credit: 1,061,708
RAC: 0
Message 4690 - Posted 10 Jan 2009 1:13:59 UTC

I keep 2 days cache on my Windows machine and it loaded up with many WU's when there was a lot of short units. Now they are taking hours instead of minutes and I won't be able to finish them before the deadlines. Should we abort overdue WU's?

J

Joined: Jan 7 09
Posts: 3
ID: 5620
Credit: 76
RAC: 0
Message 4692 - Posted 11 Jan 2009 21:11:10 UTC
Last modified: 11 Jan 2009 21:11:37 UTC

short work units? Good joke.
17 hours 13 mins in with a 3 ghz processor focusing on it, and it's at 26.2%. This is less of a complaint than a point i'm trying to make, The deadline should let me finish it, but i'm hearing about 10 minute work units and I find it strange that this one will need 60+ hours.

P . P . L .
Avatar

Joined: Oct 20 08
Posts: 69
ID: 2725
Credit: 1,000,979
RAC: 0
Message 4693 - Posted 12 Jan 2009 5:16:41 UTC - in response to Message ID 4692 .
Last modified: 12 Jan 2009 5:21:49 UTC

short work units? Good joke.
17 hours 13 mins in with a 3 ghz processor focusing on it, and it's at 26.2%. This is less of a complaint than a point i'm trying to make, The deadline should let me finish it, but i'm hearing about 10 minute work units and I find it strange that this one will need 60+ hours.


Hi J.

I've noticed that too, some finish in a couple of minutes others take hours

i think it's just the luck of the draw, good or otherwise.

I had a look at your computer, if you want it to get through the tasks faster you can switch off H/T. I did and it does make a differance trust me. ;)

pete.
____________


J

Joined: Jan 7 09
Posts: 3
ID: 5620
Credit: 76
RAC: 0
Message 4694 - Posted 12 Jan 2009 12:28:10 UTC
Last modified: 12 Jan 2009 12:51:14 UTC

Why did BOINC just toss the work unit? I must have missed the message because it was earlier, but why did it toss all that work? I could have easily finished that by the 23rd. :@ Also, could this make a difference?
1/12/2009 7:23:03 AM|World Community Grid|Message from server: (won't finish in time) Computer on 47.4% of time, BOINC on 99.5% of that
It says that's from the world community grid server but it's work unit wasn't tossed, docking@home's work unit was <.>

Low_Gap

Joined: Oct 6 08
Posts: 2
ID: 2131
Credit: 78,798
RAC: 0
Message 4741 - Posted 17 Jan 2009 18:52:59 UTC - in response to Message ID 4693 .

I had a look at your computer, if you want it to get through the tasks faster you can switch off H/T. I did and it does make a differance trust me. ;)


What is the 'H/T' switch that you are referring to?
Profile Cori
Volunteer tester
Avatar

Joined: Sep 13 06
Posts: 161
ID: 90
Credit: 5,817
RAC: 0
Message 4742 - Posted 17 Jan 2009 19:14:22 UTC - in response to Message ID 4741 .

I had a look at your computer, if you want it to get through the tasks faster you can switch off H/T. I did and it does make a differance trust me. ;)


What is the 'H/T' switch that you are referring to?

Sounds like Hyperthreading? ;-)
____________
Bribe me with Lasagna!! :-)
P . P . L .
Avatar

Joined: Oct 20 08
Posts: 69
ID: 2725
Credit: 1,000,979
RAC: 0
Message 4744 - Posted 17 Jan 2009 20:34:50 UTC

Hi.

Yes Hyperthreading sorry i forget sometimes, to use whole words.

It makes a big differance to the the time tasks take.

pete.

____________


Profile Cori
Volunteer tester
Avatar

Joined: Sep 13 06
Posts: 161
ID: 90
Credit: 5,817
RAC: 0
Message 4745 - Posted 17 Jan 2009 20:46:13 UTC - in response to Message ID 4744 .
Last modified: 17 Jan 2009 20:47:10 UTC

Hi.

Yes Hyperthreading sorry i forget sometimes, to use whole words.

It makes a big differance to the the time tasks take.

pete.

Yes, on my oldish P4 lappy I have noticed differences in WU lengths, too.
But with HT enabled you should still get more work done because two slightly slower WUs are still crunched faster (in parallel) than two WUs in a row with HT disabled. ;-)
____________
Bribe me with Lasagna!! :-)
Low_Gap

Joined: Oct 6 08
Posts: 2
ID: 2131
Credit: 78,798
RAC: 0
Message 4750 - Posted 18 Jan 2009 7:06:32 UTC - in response to Message ID 4744 .

Hi.

Yes Hyperthreading sorry i forget sometimes, to use whole words.

It makes a big differance to the the time tasks take.

pete.



Thanks. That makes sense. Something to experiment with.
Odd-Rod

Joined: Sep 3 08
Posts: 2
ID: 485
Credit: 27,955
RAC: 0
Message 4759 - Posted 19 Jan 2009 8:28:59 UTC - in response to Message ID 4745 .

Hi.

Yes Hyperthreading sorry i forget sometimes, to use whole words.

It makes a big differance to the the time tasks take.

pete.

Yes, on my oldish P4 lappy I have noticed differences in WU lengths, too.
But with HT enabled you should still get more work done because two slightly slower WUs are still crunched faster (in parallel) than two WUs in a row with HT disabled. ;-)


I agree, Cori. My P4 definitely has a faster total throughput with HT on.
But, if the 2 WUs being crunched are memory intensive enough to cause a lot of swapfile activity, then it can actually be quicker with HT off. So it's not a clear cut choice, I'm afraid.
Regards
Rod

Message boards : Number crunching : Very short WUs

Database Error
: The MySQL server is running with the --read-only option so it cannot execute this statement
array(3) {
  [0]=>
  array(7) {
    ["file"]=>
    string(47) "/boinc/projects/docking/html_v2/inc/db_conn.inc"
    ["line"]=>
    int(97)
    ["function"]=>
    string(8) "do_query"
    ["class"]=>
    string(6) "DbConn"
    ["object"]=>
    object(DbConn)#21 (2) {
      ["db_conn"]=>
      resource(120) of type (mysql link persistent)
      ["db_name"]=>
      string(7) "docking"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(1) {
      [0]=>
      &string(51) "update DBNAME.thread set views=views+1 where id=367"
    }
  }
  [1]=>
  array(7) {
    ["file"]=>
    string(48) "/boinc/projects/docking/html_v2/inc/forum_db.inc"
    ["line"]=>
    int(60)
    ["function"]=>
    string(6) "update"
    ["class"]=>
    string(6) "DbConn"
    ["object"]=>
    object(DbConn)#21 (2) {
      ["db_conn"]=>
      resource(120) of type (mysql link persistent)
      ["db_name"]=>
      string(7) "docking"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(3) {
      [0]=>
      object(BoincThread)#3 (16) {
        ["id"]=>
        string(3) "367"
        ["forum"]=>
        string(1) "2"
        ["owner"]=>
        string(2) "79"
        ["status"]=>
        string(1) "0"
        ["title"]=>
        string(14) "Very short WUs"
        ["timestamp"]=>
        string(10) "1232353739"
        ["views"]=>
        string(3) "547"
        ["replies"]=>
        string(2) "15"
        ["activity"]=>
        string(21) "8.860974582023999e-94"
        ["sufferers"]=>
        string(1) "0"
        ["score"]=>
        string(1) "0"
        ["votes"]=>
        string(1) "0"
        ["create_time"]=>
        string(10) "1226073605"
        ["hidden"]=>
        string(1) "0"
        ["sticky"]=>
        string(1) "0"
        ["locked"]=>
        string(1) "0"
      }
      [1]=>
      &string(6) "thread"
      [2]=>
      &string(13) "views=views+1"
    }
  }
  [2]=>
  array(7) {
    ["file"]=>
    string(63) "/boinc/projects/docking/html_v2/user/community/forum/thread.php"
    ["line"]=>
    int(184)
    ["function"]=>
    string(6) "update"
    ["class"]=>
    string(11) "BoincThread"
    ["object"]=>
    object(BoincThread)#3 (16) {
      ["id"]=>
      string(3) "367"
      ["forum"]=>
      string(1) "2"
      ["owner"]=>
      string(2) "79"
      ["status"]=>
      string(1) "0"
      ["title"]=>
      string(14) "Very short WUs"
      ["timestamp"]=>
      string(10) "1232353739"
      ["views"]=>
      string(3) "547"
      ["replies"]=>
      string(2) "15"
      ["activity"]=>
      string(21) "8.860974582023999e-94"
      ["sufferers"]=>
      string(1) "0"
      ["score"]=>
      string(1) "0"
      ["votes"]=>
      string(1) "0"
      ["create_time"]=>
      string(10) "1226073605"
      ["hidden"]=>
      string(1) "0"
      ["sticky"]=>
      string(1) "0"
      ["locked"]=>
      string(1) "0"
    }
    ["type"]=>
    string(2) "->"
    ["args"]=>
    array(1) {
      [0]=>
      &string(13) "views=views+1"
    }
  }
}
query: update docking.thread set views=views+1 where id=367