Counting Credits

juan BFP
juan BFP
Joined: 18 Nov 11
Posts: 839
Credit: 421,443,712
RAC: 0

RE: RE: Hi Friends Could

Quote:
Quote:

Hi Friends

Could be just curiosity, but some one could enlight me and explain whow a Host created an 12 Nov 2012 12:40:51 UTC with a RAC 29,585.14 have 2,009,475 in Total Credits? Thats is > 143K/day just with one CAL AMD Radeon HD 79x0

Some type of warp credit engine? And please, i don´t ask about the credits, i´m asking beceause i´m curious to learn whow to configurate a host, with just 1 GPU, to do so much science in so few days.


RAC is not actually a daily average, except possibly in the long term if processing is a constant. If there are rapid changes to processing, in this case upwards, then the RAC has 'lag'. I forget the exact formula but it is roughly a rolling average with 'decay' or 'accumulation' semantics.

Cheers, Mike.


Mike

I know well about the misteries of the RAC, but that´s not the point. I´m talking about total credit production per day, that´s is a plain number directly derivated to the host performance.

Forget the RAC, If you see this host for example:

http://einsteinathome.org/host/6081833

The created date was Nov-12 last contact Nov-17 total credit >2MM, an incredible 400K/day with only 1 GPU.

To compare, one of my hosts, created on Nov 18, with 2x690 running 24/7 just produce close to 300K of total credits in almost 8 days, close to 40K/day, 10X less! but my hosts have 2 Top off the class GPUs. So the production on the first host must be close to 20X bigger.

That´s why i´m curious to know how to configurate a host to do that quantity of job, imagine the increase of production in my fleet if i could replicate the configuration?

lHj2ixL.jpg

 

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,550
Credit: 289,177,010
RAC: 93,205

RE: I know well about the

Quote:
I know well about the misteries of the RAC, but that´s not the point.......configuration?


Doh! My apologies, late night. :-)

Why don't you PM him with that and/or invite him/her here to comment ?

Hmmm ... there's GW aborts everywhere on that host.

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

juan BFP
juan BFP
Joined: 18 Nov 11
Posts: 839
Credit: 421,443,712
RAC: 0

RE: RE: I know well about

Quote:
Quote:
I know well about the misteries of the RAC, but that´s not the point.......configuration?

Doh! My apologies, late night. :-)

Why don't you PM him with that and/or invite him/her here to comment ?

Hmmm ... there's GW aborts everywhere on that host.

Cheers, Mike.


Will think about PM and see what i get.

I just take that host as an example, belive that was something more generic, something well known by the comunity and easy to share/learn.

I see your point, loot of errors, but on the other hand, error does not increase the production or i´m wrong? A hidden bug? A backdoor open? I´m so new in E@H, to much things to learn, to say something.

Thanks again for your help.

lHj2ixL.jpg

 

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,550
Credit: 289,177,010
RAC: 93,205

RE: Will think about PM and

Quote:

Will think about PM and see what i get.

I just take that host as an example, belive that was something more generic, something well known by the comunity and easy to share/learn.

Now i see your point, loot of errors, but on the other hand, error does not increase the production or i´m wrong?

Thanks again for your help.


Right. Put another way, that host is marked as doing ~3000 seconds per WU for pure pulsar work @ 500 credits per throw, BUT for 140K credits/day that's ~ 300 WU's which is ~12 WU's per hour or ~300 seconds each ?

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

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,142
Credit: 2,793,941,558
RAC: 699,444

Host 6081833 may have been

Host 6081833 may have been created on 12 November, but it has task records dating back to 21 October - check task 315191527 quickly, before it gets purged.

Quote:
Name p2030.20111015.G175.38-03.19.N.b6s0g0.00000_24_0
Workunit 135186601
Created 21 Oct 2012 13:29:12 UTC
Sent 21 Oct 2012 15:51:15 UTC
Received 21 Oct 2012 19:37:41 UTC


That can happen if multiple host records are merged, or even if a host is detached then reattached.

If you get into a PM conversation with him, can you explain how to disable CPU work fetch via preferences, please? Aborting work after it arrives is very wasteful of project resources.

juan BFP
juan BFP
Joined: 18 Nov 11
Posts: 839
Credit: 421,443,712
RAC: 0

RE: Host 6081833 may have

Quote:

Host 6081833 may have been created on 12 November, but it has task records dating back to 21 October - check task 315191527 quickly, before it gets purged.

Quote:
Name p2030.20111015.G175.38-03.19.N.b6s0g0.00000_24_0
Workunit 135186601
Created 21 Oct 2012 13:29:12 UTC
Sent 21 Oct 2012 15:51:15 UTC
Received 21 Oct 2012 19:37:41 UTC

That can happen if multiple host records are merged, or even if a host is detached then reattached.

If you get into a PM conversation with him, can you explain how to disable CPU work fetch via preferences, please? Aborting work after it arrives is very wasteful of project resources.

That info is new for me, another thing to learn. Thanks Richards for your explanation it´s allways a pleasure learn something new with your gidance.

Now is explained, was not 2MM in just few 5 days, i allways belive the sistem takes the oldest creation date when merge 2 hosts, now i´m know i´m wrong.

I agree with you, abort CPU WU after DL is a waste of resources, I could try to PM the owner of the host about aborts but i realy belive a PM from one of the admins of the project/forum will be most gently and efective.

I even know who is the owner of the host, just take that randomly because i see it´s high production and a similar creation time of mine, and i was thinking, what i miss/do wrong in the configuration on my host that give so big diference in performance?, nothing else in special.

It´s your choice, if you thing if i could help with that, just ask. I´m allways happy to try to help.

lHj2ixL.jpg

 

NICHOLAS BUCCINA
NICHOLAS BUCCINA
Joined: 30 Apr 09
Posts: 1
Credit: 33,225
RAC: 0

Sorry to hear aabout battery

Sorry to hear aabout battery problems, however, I have not recived any credits
since 11/2/2012! How come? I've been a member since early 2007 via Boinc/Synergy. Can someone help!
Thanks,
Nicholas Buccina

Claggy
Claggy
Joined: 29 Dec 06
Posts: 560
Credit: 2,695,216
RAC: 90

RE: Sorry to hear aabout

Quote:
Sorry to hear aabout battery problems, however, I have not recived any credits
since 11/2/2012! How come? I've been a member since early 2007 via Boinc/Synergy.


Your host hasn't got any tasks, hasn't done any recently, and isn't asking for any eithier:

Quote:
2012-11-26 21:21:29.8457 [PID=22160] Request: [USER#xxxxx] [HOST#4188766] [IP xxx.xxx.xxx.96] client 7.0.25
2012-11-26 21:21:29.8739 [PID=22160] [debug] [HOST#4188766] Resetting nresults_today
2012-11-26 21:21:29.8740 [PID=22160] running_frac=0.061736; setting to 0.1
2012-11-26 21:21:29.8740 [PID=22160] [send] effective_ncpus 4 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2012-11-26 21:21:29.8740 [PID=22160] [send] effective_ngpus 0 max_jobs_on_host_gpu 999999
2012-11-26 21:21:29.8740 [PID=22160] [send] Not using matchmaker scheduling; Not using EDF sim
2012-11-26 21:21:29.8740 [PID=22160] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2012-11-26 21:21:29.8741 [PID=22160] [send] work_req_seconds: 0.00 secs
2012-11-26 21:21:29.8741 [PID=22160] [send] available disk 7.94 GB, work_buf_min 8640
2012-11-26 21:21:29.8741 [PID=22160] [send] active_frac 0.994750 on_frac 0.062062 DCF 2.391982
2012-11-26 21:21:29.8766 [PID=22160] Sending reply to [HOST#4188766]: 0 results, delay req 60.00
2012-11-26 21:21:29.8768 [PID=22160] Scheduler ran 0.038 seconds

Have you managed to set No New Tasks?, or is Boinc's cache full up with work from another project?

Claggy

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,852
Credit: 111,038,410,697
RAC: 34,848,037

Hi Juan, I've followed all

Hi Juan,

I've followed all your messages since you and your friends arrived with your 'firepower' recently. I'm very glad that you chose this project as the recipient of your resources and I must say I was most impressed with your thoughtfulness when you took the trouble to advise of possible impacts on the project. It will be a sad day when you all leave so we can only hope it takes Seti a looooong time to get themselves back on track :-).

It's been very refreshing to see your obvious enthusiasm for what you are doing. The message boards are much better for your questions and contributions. When you do decide to leave, I'm sure we will all be sad to see you go. So a belated welcome from me to you and your friends. We sincerely hope you enjoy your stay here.

Quote:
Now is explained, was not 2MM in just few 5 days, i allways belive the sistem takes the oldest creation date when merge 2 hosts, now i´m know i´m wrong.


The main reason for responding to you right now was to add a little to what Richard has already pointed out. Merging always merges the older hostID(s) into the most recent one and that has the obvious side-effect of losing the initial join date for the host being merged and also losing the numerically lower hostID as well. So, if a machine loses its HDD, it would be normal to replace the drive and reinstall the OS and BOINC, etc, and reattach to the project, receiving a new hostID in the process. A subsequent 'merge' then joins all the credit allocated to the two different hostIDs but the original join date and hostID are lost.

There is an 'easy' way to solve this problem for people who are comfortable with editing the state file (client_state.xml). No, you don't even have to recover the state file on the defunct disk if it's inaccessible. It's quite easy to create a state file 'template' with minimal information and let BOINC fill in all the rest. If you want to preserve your hosts identity after an event like a disk failure, you just need to 'seed' the appropriate section of a replacement state file with two critical bits of information

nnnnnn
xxxx+1.


These values are available from the host details page on the website for the old defunct host. Of course, if the old state file is accessible you can just use it and avoid any editing at all.

I have a lot of upgraded hosts - modern CPUs and mobos but using HDDs that go back to 2001-2004 era. I regularly use a variant of this technique when upgrading a CPU/mobo. I make the new host keep the ID of the one it is replacing. Consider hostID 63256 as an example. It was a single core Athlon XP running Windows - one of the very first machines I attached to the project when the project opened its doors in early 2005. For sentimental reasons, it was still running until quite recently but it is now permanently shut down and its identity has been transferred to one of my new GPU crunchers running Linux. I get quite a kick out of seeing the old, numerically very low hostIDs living on :-).

Cheers,
Gary.

juan BFP
juan BFP
Joined: 18 Nov 11
Posts: 839
Credit: 421,443,712
RAC: 0

Thanks for the info, an

Thanks for the info, an interest tip, will put in my "box of new knowledges". As you could see i´m a very courious men. I allways say "living and learning". I´m sure i will use that tip, we normaly make a restauration/upgrade on all our fleet each 3 years. This is sure one of the 100´s features hidden/missing on a corner of some page of the Boinc manual, we the end user´s will never find/noticed or even know it exist.

I realy apreciate the warm welcome, and you could notice that easely because i get so many friends with me.

I could not be one of the top BOINC crunches, but i know the impact of a combined fleet of few off us could do to any project, thats is the reason i ask before joint, don´t want to make any negative impact on project resources.

We are here to help not to cause any trouble, i´m very glad and belive i make the right choice choosing Einstain as our secondary/backup project.

In my thoughts, unless we were invited to leave, i will remain crunching Einstein even when SETI returns, ok not with all the ships but allways doing some job. Can´t speak for the others, but i know some of them will do the same.

Just need to learn how to do some optimizations for run both projects simultaneusly because their particulatities (most because the diferent rate of the CPU/GPU usage) and find the optimal point to use on each one of my hosts.

For those how could take wrong the words "secondary/backup project" please don´t do that, each man have his dreams, try to help to fild our little green man is one of mine´s, so SETI is a way to do that, but i have another dream, be happy and try to help others, and i find that on E@H, so i use this words just because i can´t find a better way to describe what i´m doing and with my bad english... i don´t spect to find better words.

Leveraging, did you think is possible, in the near future, to add the total numbers of WU with pending credits as i explain before? Thats one thing i realy miss from SETI Boinc, a way to know easely the total production of the entire fleet/host, that number gives us a fast way to know if all ships are working fine without the need to go ship by ship asking if everything is OK. Just the number of WU, don´t whant to overload any server database.

Thanks for your time a patience.

lHj2ixL.jpg

 

Comment viewing options

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