much I/O, long run times
14 Apr 2011 9:33:36 UTC
Topic 195752
(moderation:
Some of my SG5 tasks have started to behave strangely. Examples from h1_1445.80_S5R4.
A typical task, run time ~27,000 secs cpu ~26,000 secs
‘Strange’ ones. They have long periods of continuous I/O, ~30 minutes in some cases. When I monitored, I/O reads and I/O Others were clocking up at a furious rate.
2 examples
runtime >40,000 secs cpu ~36,000 secs
E@h has plenty of spare disk space. Happens when m/c is being used and when otherwise idle.
Not a problem, just curious.
Mike
much I/O, long run times
)
Drat! picked a wrog'n (I have 2 m/cs). Ignore previous links
A typical task, run time ~40,000 secs cpu ~36,000 secs
226594608
One of the ‘strange’ ones. They have long periods of continuous I/O, ~30 minutes in some cases. When I monitored, I/O reads and I/O Others were clocking up at a furious rate.
2 examples
runtime ~40,000 secs cpu ~36,000 secs
226659053
Mike
Your host has just 1 GB RAM,
)
Your host has just 1 GB RAM, so depending on other tasks running on this computer, it might be swapping to disk causing the excessive I/O.
An SG5 task will use 256 MB RAM and afaik this is initialized before a task starts to run. Your desktop uses RAM, many open browser windows/tabs can use a lot of RAM. etc.
You can see that:
[pre]2011-04-12 22:54:18.9011 (2688) [normal]: Reading input data ... done.
% --- GPS reference time = 847063082.5000 , GPS data mid time = 847063082.5000
% --- Setup, N = 205, T = 90000s, Tobs = 56435059s, gammaRefine = 1399.000000
2011-04-12 23:34:32.7267 (2688) [normal]: INFO: No checkpoint h1_1457.85_S5R4__410_S5GC1HFa_0_0.cpt found - starting from scratch
% --- Cpt:0, total:834, sky:1/139, f1dot:1/6
2011-04-12 23:34:32.7423 (2688) [normal]: 1/1[/pre]
There is a time gap of 40 min before the task is actually running.
Hi ziegenmelker, There isn't
)
Hi ziegenmelker,
There isn't (wasn't) ram contention. This isn't a heavily used box. Certainly not for 1/2 hour at 23:00, when I wasn't using the machine.
I first noticed something odd when I heard the disk clicking continuously (it's an old box) when I walked past. Dug around a bit, only E@h runnning, about 20 mins into a wu. So rebooted and let the wu start again. After 20-30 mins of the same, with 0% progress, I assumed that it was a rogue wu and killed it.
Then kept an eye on later wus and saw that they took 2 - 3 hours longer than usual for the same cpu usage.
Heavy swapping/paging would fit the symptoms. I let BOINC use half the ram (only run E@h). I thought that maybe there's something in this dataset that causes E@h to build huge resident tables etc.
Mike
I have never faced a defect
)
I have never faced a defect Wu so far, so this shouldn't be the reason. But your network card, your Internet connection, your router etc can cause defective downloads. But I doubt that is your problem.
But if your disk clicks(I have one too), it might be defect. From mine I do know, cause I enabled S.M.A.R.T in the bios.
So you might download a tool from the manufacturer of the disk and do a check for defects.
Long time ago there was an option to do a low level format of disks and it was even possible to read the defects table. At that time discs were really expensive and it made some sense. With the low budget discs today I doubt there is a way to do the same. Low level formating entered the defects to the defects table so they were not used again. But it's always a good advice to change discs that start to click irregular.
So my advice in short. Download the tool for your disk and check it. Enable S.M.A.R.T support in the bios, causing read/write errors come to your attention.
HTH
[edit]Any maleware scanner accessing the BOINC data dir? Look up the amount of RAM BOINC can use, if the computer is in use. You might also try "leave the application in memory while suspended".[/edit]
RE: (..snip..) Long time
)
i think some of the information your looking for is actually handled by s.m.a.r.t. i.e. realocated sectors (essentially the defects table)
i use speedfan to monitor my system temps, one of its features is the ability to read the s.m.a.r.t. data and perform a short and long s.m.a.r.t. test. and to all so generate a report, like this one for my hard drive speed fan hard drive analysis report
example "NOTE : your hard disk has 22 reallocated sectors."
theres a few stand alone applications to do the same, such as smartmon tools. but trying the drive makers software first is a good idea, before trying other software.
even on my E5200 with 3GB's ram it takes 6 minutes to start a WU
[pre]
2011-04-16 19:08:37.7189 (3208) [normal]: Reading input data ... done.
% --- GPS reference time = 847063082.5000 , GPS data mid time = 847063082.5000
% --- Setup, N = 205, T = 90000s, Tobs = 56435059s, gammaRefine = 1399.000000
2011-04-16 19:14:04.2539 (3208) [normal]: INFO: No checkpoint h1_1453.60_S5R4__886_S5GC1HFa_3_0.cpt found - starting from scratch
% --- Cpt:0, total:834, sky:1/139, f1dot:1/6
2011-04-16 19:14:04.2629 (3208) [normal]: 1/1
[/pre]
but on my older system P4 2.0 ghz with 1GB of ram it takes 1 hour 25 minutes
[pre]
2011-04-16 12:13:30.8986 (3812) [normal]: Reading input data ... done.
% --- GPS reference time = 847063082.5000 , GPS data mid time = 847063082.5000
% --- Setup, N = 205, T = 90000s, Tobs = 56435059s, gammaRefine = 1399.000000
2011-04-16 13:38:35.6173 (3812) [normal]: INFO: No checkpoint h1_1492.95_S5R4__779_S5GC1HFa_1_0.cpt found - starting from scratch
% --- Cpt:0, total:834, sky:1/139, f1dot:1/6
2011-04-16 13:38:35.6486 (3812) [normal]: 1/1
[/pre]
thats assuming ive read that info right and done the math correctly. both systems have the same antivirus software with boinc excluded from scans. i know its an apples to oranges comparison since both systems arent identicle, but it might give some insight.
seeing without seeing is something the blind learn to do, and seeing beyond vision can be a gift.
RE: Hi ziegenmelker, There
)
Are you sure about this? w7's footprint is about 400MB, and at least in systems with more ram is tries to keep about 25% free. If the number shown by task manager on the performance tab for Commit is larger than the physical memory used percentage shown in the status bar (or IIRC the number shown in the memory load bar) your system is swapping stuff to disk. If your computer has an IGP instead of a standalone GPU the situation is worse since it's taking a chunk of your memory away from normal use to be able to display graphics.
Should have mentioned. This
)
Should have mentioned. This is an XP box, no GPU. I ran speedfan utility a couple of days ago, no problems. As I said, it's an old box. The disk disk has always clicked faintly, they did that in the good ol' days. In comparison to my first, an early 486, it's close on silent.
Ah well, I guess this is one for the mystery file.
Mike