Message boards : Number crunching : I think we should restrict work units
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 11 · Next

AuthorMessage
Osku87

Send message
Joined: 2 Nov 05
Posts: 21
Credit: 105,075
RAC: 0
Message 13943 - Posted: 11 Jun 2006, 17:56:20 UTC

1fast6: Haven't you read that the projects big idea for the moment is to NOT release work all the time. Is it any sense to waste computer time for meaningless crunching?
ID: 13943 · Report as offensive     Reply Quote
Philip Martin Kryder

Send message
Joined: 21 May 06
Posts: 73
Credit: 8,710
RAC: 0
Message 13945 - Posted: 11 Jun 2006, 18:45:45 UTC - in response to Message 13940.  

...
its painful to watch the servers wait for a timeout for less than 30 workunits to be returned...


do we really know that the servers are waiting for stragglers?

Or, could they be waiting for their own analysis to complete in order to build the next work units based on the work completed.


ID: 13945 · Report as offensive     Reply Quote
Profile 1fast6

Send message
Joined: 2 Jun 06
Posts: 5
Credit: 245,858
RAC: 0
Message 13948 - Posted: 11 Jun 2006, 19:10:42 UTC - in response to Message 13945.  

1fast6: Haven't you read that the projects big idea for the moment is to NOT release work all the time. Is it any sense to waste computer time for meaningless crunching?


I guess it depends on your definition of meaningless crunching... all I know is I could have done SOMETHING to contribute this past week... ok, so how about, continue to release workunits that remain in "active status", until the results are validated...


its painful to watch the servers wait for a timeout for less than 30 workunits to be returned...


do we really know that the servers are waiting for stragglers?

Or, could they be waiting for their own analysis to complete in order to build the next work units based on the work completed.

[/quote]

heck,I dunno if they're stragglers, but the server status has looked pretty much like this all weekend...
Server Status

Up, out of work
22 workunits in progress
40 concurrent connections


I'm just assuming there a just a few workunits hung up in limbo somewhere, just waiting for the third submission to validate the results... but maybe I'm wrong...
ID: 13948 · Report as offensive     Reply Quote
Philip Martin Kryder

Send message
Joined: 21 May 06
Posts: 73
Credit: 8,710
RAC: 0
Message 13951 - Posted: 11 Jun 2006, 20:18:42 UTC - in response to Message 13948.  

... but maybe I'm wrong...


Perhaps.
We all run that risk - particularly when we speculate without data.

I've started another thread asking if there is any way to view the WUs.

Is it possible that those WUs have really reached Quorum, and that the fact that there are no WUs available is that there is some other asynchronous work (perhaps being done within CERN - perhaps even by **humans**) that needs to complete before more work can be sent out?

We really don't know why there is no work, or do we?



ID: 13951 · Report as offensive     Reply Quote
Profile David Lahr

Send message
Joined: 27 Dec 05
Posts: 7
Credit: 461,367
RAC: 0
Message 13954 - Posted: 12 Jun 2006, 0:03:48 UTC - in response to Message 13951.  

The way I see it, there's a way to "exploit" the system to make sure you get a lions share of the work. However, this wasn't kept a secret, so if anyone wants to they can - hence there's no secret advantage being used against people.

As someone said, if LHC want's results returned faster, they can modify the BOINC server settings, so no worries there.

I think the motive for doing it is stupid (e.g. "it's more important for me to score points than for the work to be shared and completed faster") but now that the secret is out, the best way to get back at the originators is to do the same yourself. Then they won't be getting a tactical advantage. And if the problem gets severe enough, LHC will do something about it, and we'll all be back on the level playing field.

ID: 13954 · Report as offensive     Reply Quote
Philip Martin Kryder

Send message
Joined: 21 May 06
Posts: 73
Credit: 8,710
RAC: 0
Message 13957 - Posted: 12 Jun 2006, 1:42:59 UTC - in response to Message 13954.  

The way I see it, there's a way to "exploit" the system to make sure you get a lions share of the work. However, this wasn't kept a secret, so if anyone wants to they can - hence there's no secret advantage being used against people.

As someone said, if LHC want's results returned faster, they can modify the BOINC server settings, so no worries there.

I think the motive for doing it is stupid (e.g. "it's more important for me to score points than for the work to be shared and completed faster") but now that the secret is out, the best way to get back at the originators is to do the same yourself. Then they won't be getting a tactical advantage. And if the problem gets severe enough, LHC will do something about it, and we'll all be back on the level playing field.


It is interesting to me that some consider raising the cache (or time to connect) to be "unfair".

But, raising the resource percentage to favor LHC by large factors is "ok" and even somewhat encouraged.

Both have the INTENT of getting more LHC work.
One is more effective than the other.

One is considered "ok".
While the other is considered "unfair" (by some)....

Is this rational?







ID: 13957 · Report as offensive     Reply Quote
Profile Keck_Komputers

Send message
Joined: 1 Sep 04
Posts: 275
Credit: 2,652,452
RAC: 0
Message 13958 - Posted: 12 Jun 2006, 2:11:51 UTC - in response to Message 13957.  

It is interesting to me that some consider raising the cache (or time to connect) to be "unfair".

But, raising the resource percentage to favor LHC by large factors is "ok" and even somewhat encouraged.

Both have the INTENT of getting more LHC work.
One is more effective than the other.

One is considered "ok".
While the other is considered "unfair" (by some)....

Is this rational?

I don't think anyone considers raising the cache size to be unfair, however it is counterproductive for this project.

Raising the resource share is productive for this project.

This project generates work as needed and when the work is needed it is needed quickly. That is one of the reasons more copies are sent out initially here than at any other production project. A large resource share will cause a host to get work when available, and concentrate on that work so that it is returned quickly. A large cache will cause work to spend more time sitting idle on a host while other work is being processed, delaying it's return to the project.

So as you can see it is very rational to encourage a high resource share and a small cache.
BOINC WIKI

BOINCing since 2002/12/8
ID: 13958 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13960 - Posted: 12 Jun 2006, 11:42:16 UTC

"I don't think anyone considers raising the cache size to be unfair, however it is counterproductive for this project"

I DO! I thinks its totally unfair...
just cuz the option is there, doesent make it right for all to change it... it is there for a group of puters on ones network/team, not really meant for the single puter user, cuz you shouldent need it if everyone wasent getting 3-5 days of cache at a time.. the work would be returned faster if more puters were able to process the units, the problem is were WAITING FOR THE GREEDY ONES TO FINISH THE UNITS THEY ALREADY HAVE, cuz they got like what 10- 15 units. having 10 units sitting waiting to be processed while others are dry is sooo stupid.

ive had 1 or 2 units of crunching in like two weeks... its always outta work cuz of you greedy ones. so how is letting a majority of your community run without units for like 6 outta 7 days a week, be productive?

and also why accept anymore user to join? if you dont have the work for the ones already here.

ID: 13960 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13961 - Posted: 12 Jun 2006, 11:47:39 UTC

"Is it possible that those WUs have really reached Quorum, and that the fact that there are no WUs available is that there is some other asynchronous work (perhaps being done within CERN - perhaps even by **humans**) that needs to complete before more work can be sent out?

We really don't know why there is no work, or do we?"

good point, it may not always be greedy ones taking all the units, there are rare times that like he said above... might be something going on we dont see.


ID: 13961 · Report as offensive     Reply Quote
Dronak
Avatar

Send message
Joined: 19 May 06
Posts: 20
Credit: 297,111
RAC: 0
Message 13963 - Posted: 12 Jun 2006, 12:27:44 UTC - in response to Message 13958.  
Last modified: 12 Jun 2006, 12:29:15 UTC

I don't think anyone considers raising the cache size to be unfair, however it is counterproductive for this project.

Raising the resource share is productive for this project.


A large cache is fair in the sense that it's not against the rules, but as you said, it's counterproductive, and at least some people think it's unfair, probably because of this. It's kind of like saying an action is not illegal, but it is immoral. You can still do it because you're not breaking any rules/laws, but not everyone is going to like it or agree with it. They can't do much about it though either.

As I understand the BOINC system, you're right about how the settings work, so I agree. A large resource share tells your computer to spend a lot of time on LHC. A large cache tells you computer to grab a lot of work for LHC. The idea of the first is to complete work for LHC faster by not splitting your time among multiple projects when LHC has work to do. Getting more work units as well is a result of the setting, but not the primary intention. So this is productive. The idea of the second is to grab a lot of work so you have it for a long period of time. There are some good reasons for doing this, but getting lots of units so you get lots of credit is not one of them. You slow down the project in this case by sitting on work that other people could have been doing. So this is counterproductive for the project.

I suppose there could also be something going on behind the scenes that we're not aware of, but would this site still list work units as in progress, not returned, if they weren't actually out in the hands of BOINC users? I suspect not, but I don't know for sure, that's why I'm asking.
ID: 13963 · Report as offensive     Reply Quote
Profile Ray Murray
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 281
Credit: 11,859,285
RAC: 1
Message 13964 - Posted: 12 Jun 2006, 13:39:18 UTC - in response to Message 13961.  

...the fact that there are no WUs available is that there is some other asynchronous work (perhaps being done within CERN - perhaps even by **humans**) that needs to complete before more work can be sent out?

At the start of the project there was always plenty of work from the initial studies but since that is complete, there is less for us to do, but more for those on site, doing the actual building. My understanding, although I am sure somebody will correct me, is that the Accelerator, all 27 kilometers of it (that's more than 16 miles for non-europeans), has to be designed, built and calibrated a section at a time. This means that they cannot send out the simulation work for section F for us to crunch before they have sections A through E physically completed and tested. The project isn't just about numbers, there are magnets to position and thousands of miles of associated cabling to lay. It is therefore no wonder that work is sporadic, while the physical catches up with the theoretical.

While work is available, I have my cache set to 3 days, knowing that my machine will complete all that work in little over 2 days. I suspend work on all my other projects so that my machine works exclusively on LHC when it can. This usually means that I only get two or rarely three batches of WUs before they run out, but I'm happy that the work is returned within a couple of days and certainly well within deadline. If that makes me a credit whore then colour me guilty. I do agree, however, that it isn't sensible to ask for 10 days work and then take 3 weeks to complete it. This clearly just causes annoyance to those who can't get any work at all. I also agree that perhaps the Account Creation should be switched off as there is clearly already a wide enough user base, and maybe shorter deadlines would limit the number of hoarded WUs, allowing them to be sent out again earlier if they timeout. The problem there is that there doesn't appear to be an Admin just now, until they can pressgang one of their students.
ID: 13964 · Report as offensive     Reply Quote
Philip Martin Kryder

Send message
Joined: 21 May 06
Posts: 73
Credit: 8,710
RAC: 0
Message 13965 - Posted: 13 Jun 2006, 2:08:23 UTC

Perhaps this is a Social or Psychological experiment and not one of Physics...

I just see the paper on "Volunteer Responses to Contrived Shortages in an Altruisic Computing Environment..."


Does anyone have any insight into how CERN intended to build and configure the LHC **before** they joined BOINC?
ID: 13965 · Report as offensive     Reply Quote
ralic

Send message
Joined: 2 Sep 04
Posts: 28
Credit: 44,344
RAC: 0
Message 13966 - Posted: 13 Jun 2006, 9:55:51 UTC - in response to Message 13951.  

I've started another thread asking if there is any way to view the WUs.

Why do I get the feeling that you'd like to see the wu's, so that you can try and associate them with a user whom you may believe can be pressured into returning them faster?

I smell a witchhunt brewing. :(

I no longer micromanage my systems, I rely on the infrastructure to work as designed. Granted, in the early days under certain conditions (quite a few actually :) the infrastructure failed to work properly. Those days are pretty much history, rendering micromanaging unnecessary.

The system has worked well enough to date. If the project team thought there was a flaw, or the scientists were clammering for answers earlier, they may be inclined to address a perceived "problem".

IMHO, impatient users are not a problem and the system doesn't look broke...Does it really need fixing?
ID: 13966 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13967 - Posted: 13 Jun 2006, 10:11:26 UTC
Last modified: 13 Jun 2006, 10:14:16 UTC

dont need to go any farther then to look at the top crunchers to see who is hoarding all the work units. kinna simple really :) no witch hunt neccessary :)
well at least there was a top users selection on seti at home so I may be wrong havent looked on here yet.

and its only a small amount of people that are changing thier caches like the 15-20 units were waiting on right now. most people just install and use, and never even come into forums or to go to change prefrences.
ID: 13967 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13968 - Posted: 13 Jun 2006, 11:43:59 UTC
Last modified: 13 Jun 2006, 11:49:43 UTC

69,624 total LHC@home users... isnt the work unit total like half that anyways? if so seems more users then even work units generated... but I cant remember the starting number of workunits?

ok got this off the LHC@home alpha site... so they have at least this many units... too bad acount creation is closed lol

Server Status

Up, 143681 workunits to crunch
38 concurrent connections

http://lhcathome-alpha.cern.ch/

ID: 13968 · Report as offensive     Reply Quote
Dronak
Avatar

Send message
Joined: 19 May 06
Posts: 20
Credit: 297,111
RAC: 0
Message 13969 - Posted: 13 Jun 2006, 12:28:32 UTC - in response to Message 13968.  

69,624 total LHC@home users... isnt the work unit total like half that anyways?


The "Work to be done!" thread has a few numbers. I think this most recent batch was around 67000 units, but the one before this was about 85000 units. So I guess it's kind of variable.

ID: 13969 · Report as offensive     Reply Quote
Profile Steve Cressman
Avatar

Send message
Joined: 28 Sep 04
Posts: 47
Credit: 6,394
RAC: 0
Message 13971 - Posted: 13 Jun 2006, 15:22:28 UTC

there is no where near that number of active users.

Users
32,582 total
11,866 active


98SE XP2500+ @ 2.1 GHz Boinc v5.8.8
ID: 13971 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13973 - Posted: 13 Jun 2006, 18:38:03 UTC

I got that total from a page that listed every boinc user total... so it may or maynot be authentic as I dont know who made it.

ID: 13973 · Report as offensive     Reply Quote
bowlingguy300

Send message
Joined: 1 Sep 04
Posts: 14
Credit: 3,857
RAC: 0
Message 13978 - Posted: 14 Jun 2006, 14:32:59 UTC

wow this is taking forever for you guys to finsh your over cached WU's...

what a joke.
ID: 13978 · Report as offensive     Reply Quote
m.mitch

Send message
Joined: 4 Sep 05
Posts: 112
Credit: 1,864,470
RAC: 0
Message 13980 - Posted: 14 Jun 2006, 17:33:19 UTC - in response to Message 13978.  
Last modified: 14 Jun 2006, 17:36:49 UTC

wow this is taking forever for you guys to finsh your over cached WU's...

what a joke.


What over cache? These ones will be lost or abandoned WU's. Redundant now as the quorums have well and truly been met. Any in that category are orphaned in the DB. Know problem and no fix yet.

I see you have your computers hidden so we can't tell what you've done, hmmm.




Click here to join the #1 Aussie Alliance on LHC.
ID: 13980 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 11 · Next

Message boards : Number crunching : I think we should restrict work units


©2024 CERN