Stats updates

[SandStar]

When are they?

I've let a pc run the project for a while and am now really looking forward of how it did.

[xilman]

Originally Posted by SandStar

When are they?

I've let a pc run the project for a while and am now really looking forward of how it did.

Good question. Without knowing which PC you mean, it's hard to tell.

The usual cause of misplaced stats is that a client fails to register itself properly for one of a few different reasons. Its work is then lumped in with all the other "unknown" contributors.

Please let me know (email to paul@nfsnet.org or by posting on the forum) the following information: machine's guid, its name, the pool from which it is getting its tasks and the username you typed in when you ran nfsnetinit to get things set up. If you are running the Windows GUI version, the first three should be displayed there. If you are running the command line client, it should display the pool; the guid and machine name can be found in the file processors/config.txt. In both cases, I assume you know your own username.

Given that information, it should be fairly straightforward for one of us admins to track down your contributions and ensure that they are ascribed to you.

Paul

[SandStar]

It worked out. I've just found myself in the stats.
But i´m still wondering when stats updates are run. It took about a full 24 hours to be listed here.
So are stats updates done every 24hrs or more often?

http://www.nfsnet.org/stats2/statsre...ch_Power_Cows&

[xilman]

Originally Posted by SandStar

It worked out. I've just found myself in the stats.
But i´m still wondering when stats updates are run. It took about a full 24 hours to be listed here.
So are stats updates done every 24hrs or more often?

http://www.nfsnet.org/stats2/statsre...ch_Power_Cows&

Under normal circumstances the stats are updated daily.

The procedure is as follows: at midnight UTC each day, the servers stop saving new results in the directory they've been using the previous day and start saving them in a new directory for the new day. After a short wait (15-20 minutes) to ensure that everything has settled down, they then bundle up all the relations in each pool and send them off to a Warehouse machine where they are accumulated until enough have been produced that the factorization can be completed. At the same time, the statistical summary for each pool is produced and that sent off to a Beancounter machine. Beancounter itself waits for a while to ensure that all the summaries have arrived, and then creates the statistics which are displayed on the web site. Under normal circumstances, then, the stats become visible at around 01:00 UTC. At the moment, this corresponds to 3am in much of Europe, 2am in the UK, 9pm the previous day in the US east coast and 6pm on the west coast .

Under abnormal circumstances, anything can happen. Severe thunderstorms in the Ottawa area appear to have caused power outages, taking out Beancounter for a while. I don't yet know the exact situation. Most likely Don will get it fixed later today when he gets in to the office and run a stats update. What I do know is that the data is safe and that the stats display is presently out of date.


Paul

[Jeff Gilchrist]

Yes, there were several thunderstorms that passed through in the past couple of days. We lost power here once for about 20 seconds but it was enough to trip anything not on a UPS. Hopefully Don's computers faired well...

Jeff.

[SandStar]

We´ll see what will happen :)

btw, are there any type of cpu's who do a significant better job at this project?
For example SOB and Folding@home do better with Intels while AMD's do better on others.

[Wacky]

Originally Posted by SandStar

We´ll see what will happen :)
btw, are there any type of cpu's who do a significant better job at this project?
For example SOB and Folding@home do better with Intels while AMD's do better on others.

On this topic, I will not claim to speak authoritatively. However, it is my belief that the FSB architecture is much more significant than the CPU.
NFS Sieving does very simple operations on a very large array of memory.
Although the implementation algorithms are optimized for L2 cache size, etc. the general performance is limited by the memory system. Instruction pipelining, etc. don't really help.

Perhaps Bob S. will elaborate on his research in this area.

I do know that, unlike some other computations, NFSieving is memory intensive and really works efficiently only when the system has enough real memory to avoid swapping.

[SandStar]

So if FSB is important than Intels would benefit from this since they usually have a higher FSB?

Does it also matter what kind of RAM you use? I know there are different speeds, I myself got a couple of P4 systems with RDRAM instead of the usual DDR.

aaah questions questions questions ;)
sry i´m just really curious.