Folding Forum

Or they could simply be shutting down for the New Year holiday. Let's hope that's all it is...

7im wrote:Or they could simply be shutting down for the New Year holiday. Let's hope that's all it is...

I feel sorry for the guy that had to babysit it over Christmas

on twitter they said they were doing soak testing. how long does something like that continue?

Probably a dumb question but...
Could such a high power folder lead to a shortage of WUs ?

Mactin wrote:Probably a dumb question but...
Could such a high power folder lead to a shortage of WUs ?

There is always the possibility that if there is an extremely rapid increase in the number of clients, it will outstrip the excess server capacity or the excess project backlog, and that would be particularly bad if the rapid increase were temporary. Apparently that isn't happening here.

The FAH project has repeatedly said they had a lot more work that they wanted to do that was possible with the number of clients they had at that point in time. They do plan for regular growth, adding more servers and more projects on a regular basis. They have stayed far enough ahead of the growth curve that they rarely have had serious WU shortages, though there have always been brief shortages in limited areas such as Bigadv, where everybody wants the same thing and a server is down.

Jorge1950 wrote:

Ivoshiee wrote:It is a pure speculation on my part, but such systems are expected to be short lived. They are likely just power running/stress testing their system and soon left us. If they continue then at much lower rate.

This statement was successful.
HPCS has finished their participation in the early hours of today.
It has left to FAH 595 million points, in 16 days of participation. It reached 46 position.

That is a crazy amount of cpu time. My hat's off to them for completing their burn in using fah rather than something pointless.

Ravage7779 wrote:That is a crazy amount of cpu time. My hat's off to them for completing their burn in using fah rather than something pointless.

I completely second that! I mean they chose to help cure diseases and solve outstanding problems in biology rather than analyse radio signals for a threatened project in the off-chance that one might be an alien signal, search for some special numbers, look for gravitational waves, try to crack some non-crucial ciphers, or help predict weather patterns, ...

I mean that's basically all the productive things they could have done. I am very impressed and pleased that they chose to run F@h instead!

Or thank them in person. Comment on the F@h tweet, or retweet, as you like...

http://twitter.com/hpcloud

Oh, look! PG fixed some SMP servers and HPCS is submitting work again.

spider220075 wrote:I noticed one strange thing about their production. Team 24h average: 31.682.054, total users: 1, User 24h average: 16.777.215
http://folding.extremeoverclocking.com/ ... =&t=213904
How could this be possible? Both numbers should have been identical. Oh, well, maybe their production caused dizziness to the stats server and cannot keep up with all these numbers

Jason has corrected this:

Not sure how long they will keep up that PPD, but I adjusted the db fields to compensate. The next update the 24hr avg should be correct.



Thanks,

Jason

7im wrote:Oh, look! PG fixed some SMP servers and HPCS is submitting work again.

Incredible HP Cloud Service (HPCS) ended all WS to servers like them. The analysis of the statistics by OS, I think that they work in Linux.

Edit: Please, anyone know where I can find the history of Client statistics by OS.
Output: http://kakaostats.com/tp.php?t=213904

To my knowledge there is no history page for the client stats page. I think it's all on third-parties, if they offer that at all. I do have one source that comes to mind, but it might be more inferior than a daily graph thing. What info do you need the history of?

JV I'm interested in everything. I actually have fun analyzing statistics.

See the impact of the death of Steve Jobs on the production of Mac OS X/Intel (about 70 Tflops native). Or the great blackout of FAH, as 160 TFLOPS native, etc.

Cost me much see the impact on the TFLOPS of HPCS. Although in PPD HPCS it represents 14%, its effect on the TFLOPS is very small. Not be as calculated statistics. I have not found any reference. Only a thread says they are not good, too bad.

But to make interesting the subject, it is necessary to have the statistics of daily production by OS. Perhaps the solution is to store myself, it every day. Currently initiated by storing the first of the month. It is not enough.

Edit: The problem of verification of dates, output of the SMP, PS3 or the GPU from Nvidia or ATI. This problem is reflected in these statistics, knowing them read.

This account is impressive by the number of active client, but each one taken individually is not that great. Each of their node folds at ~5500 PPD, which is basically a 2 GHz quad core ...

Anyway, it's great that such a cluster uses useful computations for their burn tests

It pleases me that my humble GT240 can do 5K ppd and match one of their 'nodes' [what would that be in practice?]
The top member in my team does 1.5 million ppd, I worked out once that was about 40 machines running flat out [guestimate].
I do find having F@H running on my kit helpful during hardware changes as it gives a good steady load to both CPU[s] and GPU.
Good to see a large group get switched on to F@H and stress test the servers - maybe PG like this too