Where would I find my "Daily Result Quota"?

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2767
Credit: 4540771105
RAC: 2151578
Topic 224292

In the F.A.Q. web page here it has a question regarding "Daily Result Quota".  Try as I might, I don't find it anywhere.  The closest thing I did find was on my computer page for #12851564 which reads: "Task duration correction factor: 2.753892", and I've shown the image below with the statement highlighted in BLUE.

Is this it?  Or is it something else?

George

Proud member of the Old Farts Association

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4700
Credit: 17544258477
RAC: 6402339

Since Einstein does not use

Since Einstein does not use the more recent and popular versions of the server software, it is more difficult to find the daily quota for a host compared to simply looking at the Details/Show tab for a host on other projects that use a modern server version.

This is an example of what Details/Show shows for an application.

Max tasks per day 1064

But on Einstein, the daily quota has to be pulled out of each server reply for each host as in this example from one of your hosts.

https://einsteinathome.org/host/12851564/log

2020-12-24 22:37:14.0586 [PID=27858]    [send] effective_ncpus 28 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2020-12-24 22:37:14.0587 [PID=27858]    [send] effective_ngpus 2 max_jobs_on_host_gpu 999999

The reply shows your Max Daily Quota for both cpu and gpu is 999999.

 

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7023374931
RAC: 1811284

The only time I can recall

The only time I can recall routinely seeing my daily result quota is the page you see if you click on the date/time link next to "last time contacted server" on the details page for a particular computer.

That page is a log of last contact including details of results returned, amount of work requested, and amount of work actually downloaded.

Here is the catch: the quota shows there only if it has been reached--then it is given as the reasons that no work is sent, or no more work is sent.

Here is another catch: the quota is not a constant of nature forever, but depends on the CPU and GPU count of the machine to start with, and is temporarily reduced in the case of errors.  It recovers to the standard value very rapidly with successfully returned results without error. (a finding of invalid when compared to another result in your quorum does not count as an error)

 

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2767
Credit: 4540771105
RAC: 2151578

Thanks Keith and ARCHAE86 for

Thanks Keith and ARCHAE86 for explaining something that is a little above my pay-grade.  I'll eventually catch on.  But...

Why would the Einstein project have this on their F.A.Q. page if they don't intend to use it:

"The 'Daily Result Quota' is normally 8 tasks (per CPU, with a 4 CPU maximum). A host can request, and will receive, up to this many tasks per day and per CPU. Each time that a host returns a failed result, or 'times out' on a result (fails to return a result by the deadline) its Daily Result Quota is reduced by one. Each time that a host returns a successful result, its Daily Result Quota is DOUBLED. Note: the Daily Result Quota is NEVER allowed to be less than one, and NEVER allowed to be larger than 8 (per CPU)."

...or at least maybe they've partially updated but not to the degree that blokes like me would actually read and look at.

What both of you are saying doesn't seem to follow suite at all.  If I have a 3950X with 16 cores/32 threads that are counting as CPU's, then 8 x 32 = 256... not "999999" as Keith said.

Is there anything else that you can explain to a 'newbie' that can possibly help me understand?

George

Proud member of the Old Farts Association

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7023374931
RAC: 1811284

George wrote:"The 'Daily

George wrote:
"The 'Daily Result Quota' is normally 8 tasks (per CPU, with a 4 CPU maximum). A host can request, and will receive, up to this many tasks per day and per CPU. Each time that a host returns a failed result, or 'times out' on a result (fails to return a result by the deadline) its Daily Result Quota is reduced by one. Each time that a host returns a successful result, its Daily Result Quota is DOUBLED. Note: the Daily Result Quota is NEVER allowed to be less than one, and NEVER allowed to be larger than 8 (per CPU)."

What is wrong with that is that it did not get updated when they added an allowance per GPU, raised the allowance per CPU, and altered the maximum number of CPUs recognized for this purpose.

Which is a lot, so far as the number is concerned, but is nothing so far as the behavior is concerned, which matches what I meant to say, and what I believe to be true.

8 per CPU per day is definitely below the current value.  4 CPUs maximum is far below the current number.

As to why they did not update the FAQ, I imagine the answer is limited resources and higher competing priorities.

The number with lots of 9s is clearly not an actual daily quota.

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2767
Credit: 4540771105
RAC: 2151578

Thanks ARCHAE86.  I am

Thanks ARCHAE86.  I am getting it... slowly but surely... So basically in a nutshell the Einstein project has changed the rules mid-stream but didn't tell the players (us).

I'll chalk this up to another learning experience and keep on plugging along.

BTW, Merry Christmas!

Thanks again.

George

Proud member of the Old Farts Association

mikey
mikey
Joined: 22 Jan 05
Posts: 11889
Credit: 1828084752
RAC: 205860

George wrote: Thanks

George wrote:

Thanks ARCHAE86.  I am getting it... slowly but surely... So basically in a nutshell the Einstein project has changed the rules mid-stream but didn't tell the players (us).

I'll chalk this up to another learning experience and keep on plugging along.

BTW, Merry Christmas!

Thanks again. 

I believe if you go thru the forums pages from when it was changed you will find a post saying things were changed and people applauding the change as it meant they could get alot more tasks per device per day. As for finding it good luck, if it were me I'd just say WOO HOO and move forward.

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7023374931
RAC: 1811284

I had another try at

I had another try at searching, and did find clear documentation of a change made in 2012 in this post by Bikeman

I'll quote the essential calculation bit:

bikeman wrote:
Max tasks per day = 32 * (number of usable CPU cores) + 160 * (number of reported GPUs) , where the number of cores is limited to max 64 and the max # of GPUs is 8.

I found an occasion a few years later on which the GPU-based limit was temporarily raised, but stated that it would be lowered again when a batch of unusually quick-running work had gone by.

Applying the stated formula to my own machines would give daily quotas as follows

Host   nCPU nGPU quota
Stoll8     4      1     288
Stoll9     6      2     512
Stoll10   6      1     352

I think the structural picture is right, but that the numbers are somewhat lower than the actual current limits in force.  In particular, I suspect the current GPU multiplier may be 192 instead of 160.  Possibly the CPU multiplier may be 40.  But those are not big differences.

If you review stated characteristics of hosts participating in Einstein, you will find a number of them "claiming" implausibly many CPU cores to be usable.  I believe most participants showing this have intentionally falsified their number of usable CPU cores in order to raise their daily quota limit high enough to service highly capable GPUs.  A Radeon VII running GRP at 2X can greatly exceed either 160 or 192 tasks/day.

Confession: I have such a falsification in my own flotilla.  The machine I call Stoll9 actually has a 4-core CPU without hyperthreading, and I falsify it to claim 6.

 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4700
Credit: 17544258477
RAC: 6402339

The use of the quoted

The use of the quoted "falsify" is not entirely appropriate.

Boinc itself provides the means to adjust the number of cores reported to a scheduler via the cc_config.xml parameter:

and here is the client configuration document's snippet about the parameter.

<ncpus>N</ncpus>

Quote:
Act as if there were N CPUs; e.g. to simulate 2 CPUs on a machine that has only 1. Zero means use the actual number of CPUs. Don't use this to limit CPU usage; use computing preferences instead.

I just am objecting to the perceived connotation that "falsify" can be interpreted as "illegal" or "dishonest" or something.

As you have used it to gain an increase in daily tasks delivered by a high producing gpu host, I see its use as entirely appropriate.

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.