BOINC is doing only LHC which I have set at 125 (12.5%)

Message boards : BOINC client : BOINC is doing only LHC which I have set at 125 (12.5%)
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1313 - Posted: 29 Nov 2005, 23:22:55 UTC

SETI is 500 (50%)
Einstein 250 (25%)
LHC 125 (12.5%)
ClimatePrediction 125 (12.5%)

Its doing only LHC and it obviously has cached work for SETI and Einstein. ClimatePrediction seems to be broken for some reason.
ID: 1313 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 1317 - Posted: 29 Nov 2005, 23:54:18 UTC

Since your avatar picture shows you as a Goa'uld, why do you need us to tell you what you want, if your are a God and all knowing?

You don't say which BOINC version you use, or which you used before you changed to it. My answer would be "I don't know" even if I did know. :D
ID: 1317 · Report as offensive
Jim K
Avatar

Send message
Joined: 8 Sep 05
Posts: 168
Message 1319 - Posted: 30 Nov 2005, 0:19:13 UTC
Last modified: 30 Nov 2005, 0:20:22 UTC

How long has it ben running LHC? LHC may have a large debt to work off and/or may be close to deadline. This is actrually a question you should ask on the LHC website forums....
BOINC Wiki
ID: 1319 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1320 - Posted: 30 Nov 2005, 0:48:58 UTC
Last modified: 30 Nov 2005, 0:51:21 UTC

It seemed to have downloaded an equal amount of work for all projects. LHC's deadline is 12/3. SETI and Einst3ein is 12/13. It has been running LHC work exclusivly all day. Client version is 5.2.13 . ClimatePrediction finally downloaded work, a very large workunit with a deadline of sometime next year. I'm running my old SETI Client to force its 50% share. I just migrated from SETI classic this morning.
ID: 1320 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1321 - Posted: 30 Nov 2005, 1:01:47 UTC - in response to Message 1320.  

So much for running SetiCommander in addition to BOINC. BOINC's low priority seems to be higher than SETI@Home's low priority and it won't let me change it in task manager for either. Access denied for BOINC, SetiCommander changes the Seti@Home client Prio back to low and you can only choose between low and normal in SetiCommander (I prefer Below-Normal or the same 'Low' between the two). I installed BOINC as a service running under the local system account.
ID: 1321 · Report as offensive
Jim K
Avatar

Send message
Joined: 8 Sep 05
Posts: 168
Message 1332 - Posted: 30 Nov 2005, 4:09:40 UTC

If you are still running Classic Seti it will take most if not all the CPU time from Boinc Seti, so you should stop Classic and remove it from your system.....
BOINC Wiki
ID: 1332 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1350 - Posted: 30 Nov 2005, 17:13:31 UTC

Its the other away around. I keep seti classic so some SETI will still get done, but it don't work because BOINC is getting all the CPU. BOINC is still doing LHC exclusively, even into today.
ID: 1350 · Report as offensive
Jim K
Avatar

Send message
Joined: 8 Sep 05
Posts: 168
Message 1355 - Posted: 30 Nov 2005, 17:39:01 UTC

One more time.
If you are still running Classic Seti it will take most if not all the CPU time from Boinc Seti, so you should stop Classic and remove it from your system.....

I am running an 840ee 3.2 with ht and LHC has been running all 4 cpus since yesterday and that is because of the short deadline.
I know you do not understand how Boinc works, but simply put, over a period, and I will use a month, the resource share will be met for all projects, that is the way it works..

My system at times will not download from Seti, Predictor, Rosetta or LHC as it has enough work to keep it busy and meet the resource share, at times I am empty of one or more of those projects but as I stated it all works out in the end......

BOINC Wiki
ID: 1355 · Report as offensive
Michael Roycraft
Avatar

Send message
Joined: 24 Nov 05
Posts: 129
United States
Message 1356 - Posted: 30 Nov 2005, 18:57:40 UTC - in response to Message 1350.  
Last modified: 30 Nov 2005, 18:58:34 UTC

Its the other away around. I keep seti classic so some SETI will still get done, but it don't work because BOINC is getting all the CPU. BOINC is still doing LHC exclusively, even into today.


Since the LHC deadline is near, the BOINC scheduler is probably running under Earliest Deadline First mode. Once it senses less of a time pinch for that project, it should start working towards balancing work for all your attached projects. One huge caveat should be mentioned - when attached to multiple projects, "micromanaging" (suspending projects, WUs,etc) will be very counter-productive, interfering with adaptation in the scheduler and requiring even more time for adjustments to be worked out. "Resetting" a project will completely futz things up.

Regards,

Michael
"The arc of history is long, but it bends toward Justice"
ID: 1356 · Report as offensive
Bill Michael

Send message
Joined: 30 Aug 05
Posts: 297
Message 1357 - Posted: 30 Nov 2005, 19:39:40 UTC - in response to Message 1356.  
Last modified: 30 Nov 2005, 19:43:51 UTC

I keep seti classic so some SETI will still get done


Since the LHC deadline is near, the BOINC scheduler is probably running under Earliest Deadline First mode.


I will emphasize this. If you are running SETI Classic at the same time as BOINC, BOINC will get very little CPU time, will think you are running an extremely slow computer, will think it cannot make it's deadlines, and will enter EDF mode and stay there forever. You will rarely finish any work by deadline, BOINC will not be able to correctly adjust it's schedules and will not download enough new work, and will never go to "round robin" scheduling (which is what you expect to happen 'by resource share'). YOU CANNOT RUN SETI CLASSIC AND BOINC ON THE SAME PC! If you do, only Classic will work, and BOINC will not.

Your description of "doing only LHC" is perfectly normal and will not affect your desire to have 50% go to SETI; it's just that this month's 12.5% to LHC will be done first, because of their short deadlines. Resource share is not "per day", it is "over the long term". If all projects were exactly equal, then it would also appear to be "per day", but LHC has extremely short deadlines and CPDN has extremely long ones, so "per day" is simply not possible.

If you wish to run Classic right now, you should "suspend" all projects in BOINC, or just exit, and when Classic stops getting work around December 15th, you can abort the wasted BOINC work units you have that are past deadline, and it will download new ones and begin to work correctly. If you leave it running until then under the current situation, it will have calculated correction factors for all projects that are so far away from reality, it will take weeks or months to adjust.

ID: 1357 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1358 - Posted: 30 Nov 2005, 20:20:09 UTC
Last modified: 30 Nov 2005, 20:25:42 UTC

BOINC is getting all the CPU, SETI Classic (via SetiCommander) is getting 0. If i set SetiCommander to run SetiClassic at 'Normal' then Seticlassic gets most, leaving about 5-10% for BOINC. I would like BOINC and SETIClassic to share nice but SetiCommander low is probably '2' and BOINC low is '3' on the system priority. The PC is a Alienware Pentium4 3.4GHZ HT.

I could care less for the deadline. Better late than not at all. Why set such a close deadline to exploit the 'earliest deadline first' mode to cheat out more CPU from clients regardless of settings? At least it doesn't seem to download any more LHC workunits. Maybe the 100 default priority number made it download too much work? I did set the cache size to 7 days (the connect every 'x' days setting) on the inital config webform. Maybe if the share priority number was added to the initial join project web form? I think it would be even better to move the CPU share priority local.

I'll just turn BOINC off until SETICommander clears out its cache. It still has 105 SETI Workunits in cache (send 16 at a time, recieve 32 at a time, 128 cache size). I told SetiCommander to not download more workunits for SETIClassic.
ID: 1358 · Report as offensive
Michael Roycraft
Avatar

Send message
Joined: 24 Nov 05
Posts: 129
United States
Message 1359 - Posted: 30 Nov 2005, 20:20:33 UTC - in response to Message 1357.  
Last modified: 30 Nov 2005, 20:21:13 UTC

I keep seti classic so some SETI will still get done


Since the LHC deadline is near, the BOINC scheduler is probably running under Earliest Deadline First mode.


I will emphasize this. If you are running SETI Classic at the same time as BOINC, BOINC will get very little CPU time, will think you are running an extremely slow computer, will think it cannot make it's deadlines, and will enter EDF mode and stay there forever. You will rarely finish any work by deadline, BOINC will not be able to correctly adjust it's schedules and will not download enough new work, and will never go to "round robin" scheduling (which is what you expect to happen 'by resource share'). YOU CANNOT RUN SETI CLASSIC AND BOINC ON THE SAME PC! If you do, only Classic will work, and BOINC will not.

Your description of "doing only LHC" is perfectly normal and will not affect your desire to have 50% go to SETI; it's just that this month's 12.5% to LHC will be done first, because of their short deadlines. Resource share is not "per day", it is "over the long term". If all projects were exactly equal, then it would also appear to be "per day", but LHC has extremely short deadlines and CPDN has extremely long ones, so "per day" is simply not possible.

If you wish to run Classic right now, you should "suspend" all projects in BOINC, or just exit, and when Classic stops getting work around December 15th, you can abort the wasted BOINC work units you have that are past deadline, and it will download new ones and begin to work correctly. If you leave it running until then under the current situation, it will have calculated correction factors for all projects that are so far away from reality, it will take weeks or months to adjust.



Bill,

I think you've painted an extremely clear picture of the situation/consequences, and I hope that more of the displaced SetiClassic folks can take heed, and benefit from your sage advice.

Thank you.

Michael
"The arc of history is long, but it bends toward Justice"
ID: 1359 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1360 - Posted: 30 Nov 2005, 20:34:39 UTC
Last modified: 30 Nov 2005, 20:58:25 UTC

OK I hope BOINC's weightings are longterm. I wanted SETI to get exclusive use of 1 CPUCore and The rest to share the 2nd (50/25/25). If SETI runs out of stuff and the server is broken then Einstein gets to monopoize a CPU while LHC/Climate shares the 2nd 50/50. Definitly looks like it don't work work that way.
ID: 1360 · Report as offensive
Michael Roycraft
Avatar

Send message
Joined: 24 Nov 05
Posts: 129
United States
Message 1361 - Posted: 30 Nov 2005, 20:36:55 UTC - in response to Message 1358.  
Last modified: 30 Nov 2005, 20:38:39 UTC

...I could care less for the deadline. Better late than not at all.


"Late" means that the quorum for the result will be greatly delayed in forming (and, on a more personal level, that you will receive no credit for the workunit).

Why set such a close deadline to exploit the 'earliest deadline first' mode to cheat out more CPU from clients regardless of settings?


Deadlines for the individual projects are established for the requirements of the project, period. I think that you are taking this adjustment on entirely too personal a level, relax.

I think it would be even better to move the CPU share priority local.


The share %s and all your other General Preferences are propagated globally, they apply to all attached projects, with the most-recently updated preferences taking effect across-the-board as the other projects come online and communicate with their individual servers.

In retrospect, it may have been more advisable for you to attach only to Seti at first, become acclimated to Boinc, and add other projects later, on a "phased" basis, instead of trying out every flavor of candy on the shelf at once.

Respects,

Michael
"The arc of history is long, but it bends toward Justice"
ID: 1361 · Report as offensive
GreatInca
Avatar

Send message
Joined: 29 Nov 05
Posts: 9
United States
Message 1363 - Posted: 30 Nov 2005, 21:17:18 UTC - in response to Message 1361.  
Last modified: 30 Nov 2005, 21:30:37 UTC

I think LHC would be much cooler if it were for crunching actual atom smashing footage (figure out the guts of the universe from bashing very little particle things). I would do LHC 500 SETI 250 Einstein 125 Climate 63 and some other one 62 if that were the case. But optimizing a topnotch particle accelerator's reliability is cool too, so i picked it as a project, but i wasn't expecting the deadline sensitive behavior.

I relize the LHC monopoly is from the tight deadline combined with the large cache setting (7 days) and the default equal priorities that you can't change upon first attatching to the project, so it downloaded too much work for the given deadline and desired priority settings and it is prioritiezing the deadline. I still say screw the dealine sensitivity. There are plenty of other non overdue workunits to get credit for, and I still would like 50% on a 2 (virtual) processor system to mean go hog a CPU just for it (same for 25% on a 4 CPU system).

Oh yeah another idea for improvement that might m ake it more optimal to run two workunits of one project over an workunits of 2 separate projects is to have one instance process two workunits on two separate threads instead of running 2 instances of the project app (sunch as sixtrack___*.exe). The OS is smart enought to put the separate intra-application threads on sepearate processors even if it is just virtual dual-ALU single processor (or both for dual-core, hyperthreaded making 4 virtual CPUs in 1). That may help L1/L2 cache utilization and make my idea of 50% prio means hog a core no longer any good.
ID: 1363 · Report as offensive
Bill Michael

Send message
Joined: 30 Aug 05
Posts: 297
Message 1366 - Posted: 30 Nov 2005, 21:27:16 UTC - in response to Message 1360.  

I wanted SETI to get exclusive use of 1 CPUCore and The rest to share the 2nd (50/25/25).


You didn't mention that you had two CPUs... that changes things a little bit.

Let me help you get (almost) to that point. In order to prevent two results from the same project running simultaneously on a regular basis, your "connect to every" setting on the website preferences (please note it doesn't matter which project - they all get "synced" to each other within a day or so) must be very low - low enough so that you only have one result from each project on your computer at a time. If you have this at the default 0.1 and have more than one per project, you'll need to lower it even further, say 0.01. (If for other reasons you have to raise this, then you may be running at some point two SETIs, or two LHCs, etc. - which is not a problem, just an annoyance if you care about the details.)

This will mostly do what you want, with the rare exception of some project being down for a bit, or out of work - that would cause it to get "out of phase", running for example Einstein and CPDN for a bit - but if you didn't touch it, it would soon be back to what you describe. BUT, because the deadlines are not all the same, ONCE IN A WHILE, the scheduler might go into EDF mode in order to finish one particular result. This has the same effect, shifting everything "out of phase". Between these two causes, you _may_ see more "SETI not running" at any given moment than you like. You can rest assured that the scheduler really will "catch up" on the SETI - which of course will mean that you will, for some period, see two SETI's running. Still, most of the time, it will be working as you want it to.

Once Classic isn't running, and BOINC knows you have a fast PC, you won't see near as much "deadline pressure" and EDF mode. Your LHC problem was caused by the CPU contention, and will likely not happen again.

Now - you have been used to running with a large cache in Classic. If you want to continue to have a bunch of work "locally", you can - but I can guarantee that you will _often_ see two results from the same project running. It will still meet your resource share requests, but it will do it as if you had two PCs. Each CPU will spend about half it's time on SETI, and there is no telling how often you'll see duplicates running. If your reason for having the large cache was not wanting to have your computer sitting idle, you have no reason to fear that any more. Given any four projects, the chances of all four being down or out of work is very slim. Given that one of them is CPDN, where one WU can last months, even if the entire internet were down, you'd at least be using one CPU on that. (Just in case the other three projects ever ARE down all at the same time, I suggest setting CPDN to "No new work" until it is getting close to completion. The last thing you want is two CPDN results running...)

The best advice I've heard given new BOINC users is "don't panic". Even when it seems like nothing is working the way you _expect_ it to work, the chances are that it's not a bug, it's that you have the wrong expectations. And by that, I don't mean that you expect more than it is capable of doing - I mean that it is "smarter than you think". Feel free to ask all the questions you want, we'll do our best to answer them. You can also look at the Wiki for the answers to almost anything!

ID: 1366 · Report as offensive
Bill Michael

Send message
Joined: 30 Aug 05
Posts: 297
Message 1367 - Posted: 30 Nov 2005, 21:31:01 UTC - in response to Message 1363.  
Last modified: 30 Nov 2005, 21:32:24 UTC

combined with the large cache setting (7 days)


OUCH! No. You can't do that. If you were running one project - okay. Running four, the "rule" is 1/4 of the shortest deadline. I don't know what LHC's is, but even SETI's 2-week deadline means your maximum cache can only be about 3-4 days. I had assumed you were running the default 0.1... running 7 makes everything 100x worse.

If you have it set to more than this, BOINC will still meet deadlines ("almost" always) and will still match your resource share... but it will spend a LOT of time in EDF mode, running one project only.

ID: 1367 · Report as offensive
Michael Roycraft
Avatar

Send message
Joined: 24 Nov 05
Posts: 129
United States
Message 1368 - Posted: 30 Nov 2005, 21:34:23 UTC
Last modified: 30 Nov 2005, 21:35:54 UTC

Bill,

A 7-day cache was mentioned a few posts ago. From what I understand, it could be quite some time before project-switching propagates any preference changes. Do you think that some judicious WU-aborting might be advisable?

(edit) OOPS, I see you caught that while I was posting
"The arc of history is long, but it bends toward Justice"
ID: 1368 · Report as offensive
Michael Roycraft
Avatar

Send message
Joined: 24 Nov 05
Posts: 129
United States
Message 1370 - Posted: 30 Nov 2005, 21:42:09 UTC
Last modified: 30 Nov 2005, 21:46:41 UTC

GreatInca,

I don't recall if I saw it here on the BOINC messageboards or whether it was on my native Einstein project boards, but I'm pretty sure that I read that there exists a way to devote one "virtual core" to a single project.

(edit) However, I don't think that there is any provision to have a virtual core select a "backup plan" in the event of a server outage if it's dedicated to one project.
"The arc of history is long, but it bends toward Justice"
ID: 1370 · Report as offensive
Bill Michael

Send message
Joined: 30 Aug 05
Posts: 297
Message 1372 - Posted: 30 Nov 2005, 21:47:23 UTC - in response to Message 1368.  

Do you think that some judicious WU-aborting might be advisable?


Not knowing how long it will take to finish the Classic work-on-hand, probably... but I'm not sure where to begin.

@GreatInca, if you'll tell us how many days you think the Classic work will take, and exactly how many BOINC results you have on hand for which projects, and what their (approximate, don't need but to the day) deadlines are, it may be a good idea (for others who are waiting on you to return results before they get credit) to "No new work" all projects, and carefully "prune" any excess work. Michael or I can help you pick which ones matter the most. Work returned after the deadline is usually discarded; no credit, no help to the science, just a waste of CPU cycles, and because it must then be reissued to someone else, any others doing the same WU are stuck waiting on them to return the work. (There are exceptions, it's complicated.)

@Michael, if "no new work" on all, changing pref on one then update it then update others will propagate in minutes... otherwise, I think it'll take 1 WU each; I would hope that CPDN does it on trickle! :-)

ID: 1372 · Report as offensive
1 · 2 · 3 · Next

Message boards : BOINC client : BOINC is doing only LHC which I have set at 125 (12.5%)

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.