Message boards : Number crunching : RAC (expavg_credit) being recalculated incorrectly by LHC
Message board moderation

To post messages, you must log in.

AuthorMessage
Toby

Send message
Joined: 1 Sep 04
Posts: 137
Credit: 1,691,526
RAC: 8
Message 7585 - Posted: 11 May 2005, 0:29:28 UTC

I think this will take an admin to answer for sure but for discussion:

It appears as though LHC is forcing RAC to be recalculated when it shouldn't be. On my stats website I base my "Last Result Recieved" on the expavg_time field in the XML. This time stamp should reflect the last time the validator granted credit to a particular user/team/host. However it would seem that LHC is forcing an update which resets expavg_time to "now". For example, if you look at this user on our team it is clear that he hasn't returned any work units in a VERY long time. However here on my stats site it reports him as having returned a work unit today. This is obviously not correct. RAC is ONLY supposed to be recalculated by the project when new credit is granted. Other than that it is up to whoever is using the stats to decay it correctly using the expavg_credit and expavg_time fields from the XML stats. I am trying to develop a new feature on my stats site which will show how many total users/teams/hosts there are and how many ACTIVE ones there are. The only thing I have to determine wether or not they are active is expavg_time and if this is constantly being updated by LHC then this feature becomes useless. All of the other projects behave correctly soo... what did you guys change?? :)
- A member of The Knights Who Say NI!
My BOINC stats site
ID: 7585 · Report as offensive     Reply Quote
Profile Keck_Komputers

Send message
Joined: 1 Sep 04
Posts: 275
Credit: 2,652,452
RAC: 0
Message 7597 - Posted: 11 May 2005, 9:36:48 UTC

Most likely they have not changed something that most other projects have changed. There used to be a deamon that ran daily? that would decay all RACs. This deamon has been turned off in most projects to reduce DB access.
BOINC WIKI

BOINCing since 2002/12/8
ID: 7597 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,576,736
RAC: 3,733
Message 7599 - Posted: 11 May 2005, 10:34:11 UTC

HM, so far, as I understood the intention of the RAC, it was designed to show the actual power of crunching. So, if a Host / a User stoppes crunching, it would be correct, if the RAC goes down until it is 0 again ...



Supporting BOINC, a great concept !
ID: 7599 · Report as offensive     Reply Quote
Profile Markku Degerholm

Send message
Joined: 3 Sep 04
Posts: 212
Credit: 4,545
RAC: 0
Message 7616 - Posted: 11 May 2005, 21:32:23 UTC - in response to Message 7585.  


> what did you guys change?? :)

Nothing. We have a bit outdated scheduler with some RAC-related issues but we are testing the new scheduler on the alpha site. Let's hope the new version does it right.

Markku Degerholm
LHC@home admin
ID: 7616 · Report as offensive     Reply Quote
Toby

Send message
Joined: 1 Sep 04
Posts: 137
Credit: 1,691,526
RAC: 8
Message 7626 - Posted: 12 May 2005, 2:08:47 UTC - in response to Message 7616.  

> > what did you guys change?? :)
>
> Nothing. We have a bit outdated scheduler with some RAC-related issues but we
> are testing the new scheduler on the alpha site. Let's hope the new version
> does it right.

Are you SURE? Maybe some mouse in the office slipped a script into CVS somewhere? :) Unless JKeck is correct... but as far as I can tell, this has never been a "feature" of BOINC. I get XML from seti every day where the RAC hasn't been recalculated since June 24th, 2004. That is shortly after seti went live. All the other projects also have users with RAC that hasn't been recalculated since shortly after they first came online. I suppose it is possible that they all ran into DB problems immediately and took a tip from Berkeley and disabled the daemon...

@Yeti: You are correct. However it can be re-calculated on the fly using the expavg_credit and expavg_time fields which is ideally the way it should be done rather than updating it in the database every night. At least that is what it looks like from where I'm sitting.
- A member of The Knights Who Say NI!
My BOINC stats site
ID: 7626 · Report as offensive     Reply Quote
Profile Keck_Komputers

Send message
Joined: 1 Sep 04
Posts: 275
Credit: 2,652,452
RAC: 0
Message 7636 - Posted: 12 May 2005, 10:03:16 UTC

>... but as far as I can tell, this has never been a "feature" of BOINC.

It was used in the beta test project. I don't think it was ever used by the public seti project, at least not after the first crash (less than a week).

I personally think this feature should be turned back on. The projects should have the accurate numbers instead of making the stats sites calculate them.
BOINC WIKI

BOINCing since 2002/12/8
ID: 7636 · Report as offensive     Reply Quote
Toby

Send message
Joined: 1 Sep 04
Posts: 137
Credit: 1,691,526
RAC: 8
Message 7643 - Posted: 12 May 2005, 12:31:50 UTC - in response to Message 7636.  

> I personally think this feature should be turned back on. The projects should
> have the accurate numbers instead of making the stats sites calculate them.

I agree it would be good for projects to show the correct RAC for people however this might mean that something needs to be added to the XML because stats sites really need to be able to distinguish between active/inactive users and teams. Hosts have the rpc_time to go by but that really isn't ideal either because a host can perform an RPC without getting/sending any work. Since many people have their hosts hidden, you can't determing a users "last rpc time" reliably either. Hmm... maybe I'll start a discussion on the stats mailing list about this later today - it has been rather dead for a while and could use a good swift kick in the rear :)
- A member of The Knights Who Say NI!
My BOINC stats site
ID: 7643 · Report as offensive     Reply Quote

Message boards : Number crunching : RAC (expavg_credit) being recalculated incorrectly by LHC


©2024 CERN