Message boards :
Number crunching :
Sequestered CPU question
Message board moderation
Author | Message |
---|---|
Mike Gelvin Send message Joined: 23 May 00 Posts: 92 Credit: 9,298,464 RAC: 0 |
I have several Intel Hyperthreaded CPUs that I run Classic on. When I do, I only use ½ of the CPU (-CPU 1 switch) and leave the other half for user use. Most work done on these systems don’t use the FPU and the users don’t notice Seti is there. I use the –CPU switch to prevent Seti from swapping back and forth between the halves. This was a result of testing to see if I could eliminate a “lag†problem where seti was sluggish giving up the processing time. The –CPU switch apparently solved all those problems. Win NT, Win 2000 and XP are the OSs being used. My question is: Is there a similar switch in BOINC to sequester activity to only ½ of the CPU and FORCE it to only use a specified half? |
Bruno G. Olsen & ESEA @ greenholt Send message Joined: 15 May 99 Posts: 875 Credit: 4,386,984 RAC: 0 |
|
Mike Gelvin Send message Joined: 23 May 00 Posts: 92 Credit: 9,298,464 RAC: 0 |
No, That I have already done. That limits the projects from using both halves at once, but doesnt eliminate the switch between them. Classic had that ability. |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
> I have several Intel Hyperthreaded CPUs that I run Classic on. When I do, I > only use ½ of the CPU (-CPU 1 switch) and leave the other half for user use. > Most work done on these systems don’t use the FPU and the users don’t notice > Seti is there. I use the –CPU switch to prevent Seti from swapping back and > forth between the halves. This was a result of testing to see if I could > eliminate a “lag†problem where seti was sluggish giving up the processing > time. The –CPU switch apparently solved all those problems. Win NT, Win 2000 > and XP are the OSs being used. My question is: Is there a similar switch in > BOINC to sequester activity to only ½ of the CPU and FORCE it to only use a > specified half? > On the 'Your Account' page you can tell BOINC only to use 1 CPU (under Generel Preferences - the default is 2). I don't know of a way to set CPU afinity though. Paul Bucks' documentaion http://boinc-doc.net/ is an excelent reference. Actually, I am surprised that you ascribe any lag to CPU utilization. The task switching process (round-robin scheduling) occurs many times a second and is pretty efficient, BOINC projects run at the lowest priority, so any user CPU needs should be satisfied virtualy instantly. Othe system resources such as memory are not as fast to react to changes in demand, any response problem is much more likely due to another resource. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> Paul Bucks' documentaion http://boinc-doc.net/ is an excelent reference. [blush] At this time there is no mechanism to set processor afinity like old SETI@Home Classic. Frankly, as has been stated, the lag due to BOINC stopping and starting is, for almost all people, unnoticable. Other settings that you can tweak include setting the delay time to several minutes so that when work is suspended it does not start again until the computer is inactive for a long period of time (5 minutes? 15? ... 2 days?) ... If it is an HT processor, set the max to 2 and let it go, if they are not doing heavy lifting it is very unlikely that they will notice it at all. I do my work on a PowerMac and have running iTunes playing music, MySQL database running to support my site development and testing, Apache, my editor with well over 200 pages loaded at all times ... and oh, by the way, BOINC is running full time (sometimes with more than two Science Applications running at the same time - Yes, I know, it should not do that ...) ... The point of this is I never lose a beat of music, see no hesitations in the activities of the system ... |
Mike Gelvin Send message Joined: 23 May 00 Posts: 92 Credit: 9,298,464 RAC: 0 |
Paul, I assume that was IMHO... my experience is quite different. I am a heavy classic user. I have fine tuned several machines, and even wrote a custom service shell that classic runs under. I am evaluating BOINC with an eye to migration, but am finding the tools provided by BOINC lacking. 1) no setiqueue like facility for centralized collection and management of several computers with one point of contact. 2) Affinity – I found this useful, I am not saying all will. What is the development protocol? I see that the source is available, but I assume someone is in charge of it? I would like a few tweaks… and I am willing to roll up my sleeves and help. |
Bruno G. Olsen & ESEA @ greenholt Send message Joined: 15 May 99 Posts: 875 Credit: 4,386,984 RAC: 0 |
> 1) no setiqueue like facility for centralized collection and management of > several computers with one point of contact. Unfortunately I can't find it right now, but I've seen a thread somewhere about some guys working on something like that. But I suspect there would be some problems to overcome due to the more strict way BOINC works in that area. I've even seen posts stating that it is impossible. |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
> > 1) no setiqueue like facility for centralized collection and management of > several computers with one point of contact. Grab a copy of BOINCView, http://boincview.amanheis.de/ It will address all your needs as far as cnetral management is concerned. > 2) Affinity – I found this useful, I am not saying all will. > > What is the development protocol? I see that the source is available, but I > assume someone is in charge of it? I would like a few tweaks… and I am > willing to roll up my sleeves and help. Join the mailing lists at the bottom of this page: http://boinc.berkeley.edu/anonymous_platform.php You'll get to see how things happen. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> Paul, I assume that was IMHO... my experience is quite different. Oh, my yes. But, I do have experience with BOINC. And, yes, many people see things differently than me. And you are free to think that I am as "looney-toons" as you might wish (I am autistic so I won't even notice) ... :) But, you asked questions, and I tried to answer. The problem is that BOINC is not as simple as SETI@Home Classic. If it was, I would not had so much to write about. > I am a heavy classic user. I have fine tuned several machines, and even wrote > a custom service shell that classic runs under. I am evaluating BOINC with an > eye to migration, but am finding the tools provided by BOINC lacking. > > 1) no setiqueue like facility for centralized collection and management of > several computers with one point of contact. > 2) Affinity – I found this useful, I am not saying all will. I never wrote a custom shell, I used SETI Spy and SETI Driver to serve my needs. But the queuing and proxy of SETI Driver are "built in", so that is not needed. For a SETI Spy replacement there is BOINC Spy, BOINC Log-x and BOINC View (which also allows management of multiple computers running BOINC). There are other programs that do similar things. Granted, some of them are a little rough, but new ones are comming out all the time. There is even a BOINC Manager replacement out that uses PHP to manage the BOINC Daemon. > What is the development protocol? I see that the source is available, but I > assume someone is in charge of it? I would like a few tweaks… and I am > willing to roll up my sleeves and help. Well, I can't answer this well. But there are mailing lists and there is the source available to modify. As far as getting your change into the BOINC Software, I don't know how anyone clears that hurdle. It is do-able as John McLeod VII and others have been working on the BOINC Daemon's Work Buffer controls to improve scheduling of work by the BOINC Daemon. first step is to, in my opinion, join the mailing lists for developers, download the source and start to get familiar with it. If you find processor affinity a "must have" feature then you can work on that for your own use, and possibly, if you can convince Dr. Anderson of its utility, the rest of us when that change is added to the baseline. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.