Windows Performance monitor goes nuts, other windows very slow
19 Dec 2005 20:24:29 UTC
Topic 190412
(moderation:
I have setup Performance Monitor to watch each of my Boinc process for %cpu time. Einstein seems to make Performance Monitor go absolutely nuts. It will leave huge gaps in it's plot and will show Einstein with more than 100% of cpu time (like 400%?! on a single processor!). If I watch task manager at the same time I see kernel time pegged at 100% cpu. I upped the priority of Performance Monitor to high, Task Manager to realtime and made sure Einstein was set to low. When the weird behvior happens I even have trouble getting Task Manager's window to popup.
My current suspicion is that Einstein is doing some system function like allocating huge amounts of memory and that is why I see kernel mode pegged. I will be reducing it's allowed virtual memory and see. At this moment it is not well behaved for background computing. No this is not the end stage of a work set so that is not what is happening.
I didn't see a reference to this is other posts. Any comments?
Windows Performance monitor goes nuts, other windows very slow
)
Well, I wouldn't go mucking about with the priorities of any task at all for starters! I'd reset those back again if I were you.
If you want to set the behaviour of BOINC's usage then go to your account and set up preferences there.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Thanks for the reply
)
Thanks for the reply Mike,
Ah yes the Gods of Windows. Only problem is I am an Iconoclast and have my own attitude. Besides I hurt my knee so bowing is out. I happen to feel if I can't get task manager to talk to me, I don't care if anything else runs or not. Second, when I am monitoring some task(s) I expect the monitor to run at a higher priority to have reasonable accuracy. I'm not a big fan of making things work by changing priorities but I won't tolerate a background compute bound process hammering my ability to type an email.
I was there and tried that and tried to imply I was going to do it again to decrease the maximum virtual memory that Einstein could use. I suspect that system calls issued at a low priority are hammering everything of higher priority. It's sort of like a Denial of Service attack from something I want to donate some time to. If there are any revelations about this out there please let me know.
Thanks, Larry
RE: Thanks for the reply
)
All you really need to do in your account preferences web page is to:
A - specify 'no' for the "Do work while computer is in use?" question.
B - set a value that suits you for the "Do work only after computer is idle for (applies only if above is 'no')" question. Say 1 minute.
C - note which usage profile ( default/home/work/school ) the above settings refer to.
D - go to "Your computers" page, and select the computer in question, go to the "Location" item at the bottom and set it to be the above profile ( from C ). Hit Update. Now the BOINC server knows it.
E - go back to BOINC in your computer and update highlight Einstein@Home on the projects tab and hit "update" on the left. This will inform your machine ( from the BOINC server ) of which usage profile you have set on the web page ( from C ).
Now when you are not using your computer ( after the time interval set in B ), it will go for BOINC at whatever amount that Windows will allocate it. Note that normal Windows scheduling according to default priority settings will manage all the device drivers system stuff etc. If you want to micro-manage that stuff then best of luck, all we'll need is your epitaph! ;-)
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Hi Mike, I set the
)
Hi Mike,
I set the Einstein schedule for very late at night only. I thought about specifying a "no" for the "Do work while computer is in use?" question. If they are using keyboard input to determine that it's probably a good choice.
I think the problem I am trying to get at is the other projects I have so far are pretty well behaved (Seti, Climate Prediction). I have run Seti for years full time on single and multiple processor machines with very little problem as long as the priority is low. Einstein is not well behaved at all. If it were an application I HAD TO RUN I would be watching all of it's system calls, file accesses, and disk and network I/O with call by call monitoring tools - probably on an isolated machine. There are things a low priority non-privileged process might do to impact the rest of the system that are probably benign but there is lots of less benign software out there that can behave in a similar way. I guess I want to know why and don't have the time right now to dig in far enough. I thought some others might have noticed it. If I get irritated enough I probably will.
Larry
P.S. I could be wrong but it doesn't appear to be BOINC's fault unless you include the setting that allows Einstein to run at all. I remind you that Einstein is able to prevent Taskmanager (at realtime priority) from running for an appreciable fraction of a minute!!!!! Let me repeat!!!!!
RE: I remind you that
)
I understand what you're saying, but haven't stepped into this thread because I have not seen anything even _close_ to what you're describing, on Windows XP or Mac, or any reports of it. Einstein for me has been one of the _best_ behaved apps. You certainly aren't memory constrained (1GB), and while your Celeron isn't going to be called "super fast", it's not a P III with Windows 98, either... and I see no reason why Win2000 should be any different than WinXP on this.
You had a very strange error on your third result:
5.2.13
Incorrect function. (0x1) - exit code 1 (0x1)
No heartbeat from core client for 31.000000 sec - exiting
That also is something I haven't seen before, at least not on Einstein; I was expecting the "graphics bug". Someone else may have an answer for you, but this definitely is an odd one.
I think the error on the
)
I think the error on the third result was when I used Task Manager to pound a stake into it's heart (End Process) to make sure it was dead. By doing that I saw an immediate change in behavior without worrying about whether a BOINC SUSPEND was working.
The key here is that the behavior is unusual compared to other Einstein users experience. I don't see why any NT variants should be any different either and speed better not have anything to do with it as that starts to sound like a timing bug.
At this stage I will probably detach and reattach. It's possible there is a problem with the Einstein image I have. I have no indication that I have any particular malware on my machine but who ever knows for sure.
I will report later if I get the same type of behavior and I will try to get some time to characterize what is going on at the system level. Until then I will let this go...
RE: You had a very strange
)
Hey I had one of those! I'd ignored it as it was the night we had a major storm, with power outages etc. ( My 'UPS' is pretty basic ).
Result ID 12275283
Computer ID 468670
5.2.13
- exit code -1073741502 (0xc0000142)
No heartbeat from core client for 1692.023724 sec - exiting
for this machine:
CPU type AuthenticAMD
Mobile AMD Athlon(tm) XP 2400+
Number of CPUs 1
Operating System Microsoft Windows XP
Home Edition, Service Pack 2, (05.01.2600.00)
Memory 447.48 MB
Cache 976.56 KB
Swap space 1058.09 MB
Total disk space 27.92 GB
Free Disk Space 2.24 GB
Measured floating point speed 1643.52 million ops/sec
Measured integer speed 2781.06 million ops/sec
This page from the BOINC Wiki discusses this.
It waited quite a long time before deciding it was dead!
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: I have setup
)
Larry:
See the new thread I started here:
http://einsteinathome.org/node/190443
I'd wager you're actually describing the same thing.
Mark
RE: RE: Larry: See the
)