Rejecting My Computer
9 Jun 2007 0:02:37 UTC
Topic 192830
(moderation:
I turned on my computer after getting home from work today and the following messages were displayed:
6/8/2007 6:33:45 PM||Starting BOINC client version 5.8.16 for windows_intelx86 6/8/2007 6:33:45 PM||log flags: task, file_xfer, sched_ops 6/8/2007 6:33:45 PM||Libraries: libcurl/7.16.0 OpenSSL/0.9.8a zlib/1.2.3 6/8/2007 6:33:45 PM||Executing as a daemon 6/8/2007 6:33:45 PM||Data directory: C:\\Program Files\\BOINC 6/8/2007 6:33:45 PM||BOINC is running as a service and as a non-system user. 6/8/2007 6:33:45 PM||No application graphics will be available. 6/8/2007 6:33:45 PM||Processor: 4 GenuineIntel Intel(R) Xeon(TM) CPU 1.80GHz [x86 Family 15 Model 2 Stepping 7] [fpu tsc sse sse2 mmx] 6/8/2007 6:33:45 PM||Memory: 1023.01 MB physical, 2.90 GB virtual 6/8/2007 6:33:45 PM||Disk: 111.73 GB total, 20.53 GB free 6/8/2007 6:33:45 PM|rosetta@home|URL: http://boinc.bakerlab.org/rosetta/; Computer ID: 103582; location: home; project prefs: default 6/8/2007 6:33:45 PM|Einstein@Home|URL: http://einstein.phys.uwm.edu/; Computer ID: 942314; location: (none); project prefs: default 6/8/2007 6:33:45 PM|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 1957487; location: (none); project prefs: default {snip} 6/8/2007 6:43:50 PM|Einstein@Home|Sending scheduler request: To fetch work 6/8/2007 6:43:50 PM|Einstein@Home|Requesting 2160 seconds of new work 6/8/2007 6:43:55 PM|Einstein@Home|Scheduler RPC succeeded [server version 509] 6/8/2007 6:43:55 PM|Einstein@Home|Message from server: No work sent 6/8/2007 6:43:55 PM|Einstein@Home|Message from server: (won't finish in time) Computer on 39.7% of time, BOINC on 90.5% of that, this project gets 25.0% of that 6/8/2007 6:43:55 PM|Einstein@Home|Deferring communication for 1 min 0 sec 6/8/2007 6:43:55 PM|Einstein@Home|Reason: requested by project {snip rest}
OK, Mr. Einstein Server, you need a head check on your "(won't finish in time)" routine. Yes, my computer is only on 39.7% of the time and Boinc is on 90.5% of that. However, you can't then project that an Einstein@home application only "gets 25.0% of that" because I have four virtual processors (two physical processors with hyperthreading turned on): "6/8/2007 6:33:45 PM||Processor: 4 GenuineIntel Intel(R) Xeon(TM) CPU 1.80GHz [x86 Family 15 Model 2 Stepping 7] [fpu tsc sse sse2 mmx]"
Mr. Einstein Server gets one dedicated CPU (25%) of my total Boinc job mix, which is over 60 hours per week. If I left my air conditioner on when I was gone I would leave the computer running, but you have no idea how much that would cost and I just can't afford that these days.
Anyway, at 60 hours per week, I can still crank out an Einstein work unit about once every 6 to 10 days, depending on where the weekend falls in the mix (the estimates for my Xeon hyperthreaded CPUs also consistently over-estimate how long the work unit will take to complete; and I compute more on weekends as I'm home more, too).
Now, because Mr. Einstein server refused to give me another work unit, my Boinc just went and got itself another Rosetta work unit, so its not like I have nothing left to do here.
But I really think you ought to fix that computation if you are going to make these sorts of decisions based upon false facts.
== Bill
Rejecting My Computer
)
Unfortunately for you not all the units are the same size, for time and credits.
The 290 credit units you have done are lower than any I have done. And the time/credits graph, as far as I have seen is linear.
So I assume the project has no low credit units available. The 500 credit units I have done, not the highest I have seen, would predictably take ~90 hours to complete on your computer, which is more than you are prepared to donate.
We badly need the application to be optimised, but I think their priority is to get cross OS validation sorted first.
Andy
If I remember correctly the
)
If I remember correctly the server will always let you have one result even if it thinks you wont finish in time. The problem with this is that you have to report the one you currently have before you are allowed an other one.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
One other factor is this is a
)
One other factor is this is a relatively new host on EAH, therefore it most likely hasn't run enough results yet to bring the RDCF into line with the hosts true performance.
Given the limited uptime, the 25 % resource share, and the increase in the deadline 'tightness' factor of the work on S5R2, I don't find it surprising BOINC balks at attempts to put an EAH result in the 'bullpen'. Remember that every hour a result spends waiting to run is effectively one hour lopped off the deadline.
As was mentioned, once the current result is reported you will most likely get another EAH result then, or shortly after that.
Alinator
RE: ... Given the limited
)
But on a 4 CPU system, and work split among only 3 projects, a resource share of 25 % is not really a limiting factor per WU, right? Because Einstein should get a CPU 100% of the time (leaving debt to other projects aside for the moment ..). The question is: does boinc take this into consideration?
CU
BRM
This is a double post, see
)
This is a double post, see this thread for a possible answer. Dave.
RE: RE: ... Given the
)
IIRC, this is one of the limitations with the current codebase. It doesn't fully take into account the performance potential of multicore hosts when it comes to scheduling matters under all scenarios. It seems to do reasonably well with up to 2 cores, but would appear to get a little overly conservative beyond that.
Alinator