| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 20
|
|
| Author |
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
I understood that all WU are sent in double to different members to be processed. So to have validation both WU's must be returned to WCG before the points are attributed.
----------------------------------------I am starting to wonder then what happens in the following "hypothetical extreme case": It happens that I have as a single member a stellar machine extremely powerful that can process a very high number of WU per day. But as we know the second copies of each WU sent to that machine goes to a member with a very slow, old machine that processes let's say one WU per week. My understanding is that I will not get the points of my WU until that member finished processing. So I will end up with hundreds or thousands of WU piling up and getting points at the rythm not of my machine but of the slowest member machine. If my reasoning is right then there is no incentive to have very powerful machines crunching. But I suppose I am wrong somewhere. Hypernova ![]() |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
If you want more or less instant credit gratification you could consider only doing FAAH & HFCC who have no redundancy, except when random verification jobs are send out.
----------------------------------------HCMD2 does a type of CPU matching to optimize something else, so that would then be your second choice. And for me and all the others who cant be bothered with PV tasks waiting on wingmen... we know that when they do validate, and they do in 99.9999%, the taste is even sweeter ;>) Oddly, I run about 1.5 day cache and have 3 only in PV... and a substantial part is RICE and they run quorum My last 4 days has 83 results validated, so 3 on 86 is 96.5% validation within the first 4 days.Then my duo had 24 validated and the quad 57 in that period... both using same amount of e-power. Thus, if you suppose you were astray in your assumptions and perceptions... start counting ![]() edit: corrected RICE quorum. HPF2 has quorum 15.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Dec 9, 2009 11:33:30 AM] |
||
|
|
gb009761
Master Cruncher Scotland Joined: Apr 6, 2005 Post Count: 3010 Status: Offline Project Badges:
|
I'd say that the worst case scenario, is that you'll have to be waiting for a maximum of 14 days for all your WU's that go into PV status to be validated. 10 days for the original WU to be timed out, and then another 4 for the follow-up WU to be processed. Yes, there is a chance that that follow-up WU may need another 'follow-up' WU being sent out, but I'd say that was very slim (taking into consideration the fact that all 'follow-up' WU's are sent to 'reliable machines').
----------------------------------------The beauty of Grid computing, is that not all of your wingmen will have the slowest machines - you may find, occasionally, that they are waiting for YOU to return yours :-D What is certain, is that, eventually, they'll get validated and you'll get your points... All it takes, is patience :-) ![]() |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
As I said it was a hypothetical case.
----------------------------------------I will have to do some statistics to check. But I am nearly sure I do not have such a high validation rate 96.5% in four days. At the moment I have the following situation: 54 WU are PV over a span of 9 days. If I take all the WU which are Valid on the same span of 9 days I have a total of 210. This means the total number of WU over 9 days is 264. I have then a validation rate that is 79.5%. That is not too bad but could be improved. I wonder what will happen with some 22 more cores kicking in at the end of this week. Hypernova ![]() |
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
As I said it was a hypothetical case. I will have to do some statistics to check. But I am nearly sure I do not have such a high validation rate 96.5% in four days. At the moment I have the following situation: 54 WU are PV over a span of 9 days. If I take all the WU which are Valid on the same span of 9 days I have a total of 210. This means the total number of WU over 9 days is 264. I have then a validation rate that is 79.5%. That is not too bad but could be improved. I wonder what will happen with some 22 more cores kicking in at the end of this week. Hypernova I might fall into your proposed situation. I have 3 machines on WCG now.. One with 24 threads and realistically as powerfull as you can't buy yet and the other 2 with 4 cores each but one of those two is a very fast Yorkfield at 4171mhz and the other 4 core is a dual socket system using essentially 2 Yonah laptop chips at a slooow 2000mhz but it intrigues me so I keep it. in none of the machines do I ever get what i'd call a serious backlog no matter what the project and as said above there are those days where you look and had X number of unexpected WU validate and you just smile. I guess what I'm trying to say to you is have fun with the project and "Don't sweat the small stuff" Merry Christmas to you and yours. ![]() |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
Hypernova,
----------------------------------------there are two other points to correct in your reasoning. The first one is a joke, the second one is serious. It happens that I have as a single member a stellar machine extremely powerful that can process a very high number of WU per day. Not possible: traditionally at WCG it is Movieman who owns that wonderful machine. So I will end up with hundreds or thousands of WU piling up and getting points at the rythm not of my machine but of the slowest member machine. Even if you are extremely unlucky and all your wingmen put you in the worst case described by In reality things are never black or white, so you will have most WUs validated in 0 to 4 days and a few others will wait longer. So, after the startup period, you will get points more or less at your rhythm, with some days above average and some below, that's all. And back to the title of your thread the only impact on performance is because (for these projects with double distribution, i.e. HCC and HCMD2) each WU has to be computed twice. But that is the price to pay for reliability for this type of projects where the no-redundancy concept is not applicable. Cheers. Jean. Edit: It was not Sekerob, it was gb009761. ---------------------------------------- [Edit 1 times, last edit by JmBoullier at Dec 9, 2009 12:32:06 PM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
At the moment I have the following situation: 54 WU are PV over a span of 9 days. If I take all the WU which are Valid on the same span of 9 days I have a total of 210. Are you taking these numbers from your Result Status pages or are you keeping a log of all your WUs separately? The reason why I ask is that WUs are removed from the Result Status pages quite fast after they have been validated (I think that currently it is two days). So, if you are counting valid WUs from the Result Status pages over a period of 9 days you will miss the majority of them. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It is more a matter of each individual's cache size. If we are going to invent hypothetical worst case scenarios then how about this one ... what if Movieman, who has one of the top three (I believe) machines in WCG
, sets his cache to be 10 days (the maximum allowed) then most people, regardless of their PC will be waiting for his returns in order to validate when he is their wingman. Yes, how ironic would it be if the P4s of the world were waiting for MM :sofa:Like already said ... don't worry, they will all validate eventually ... we applaud your core count Realistically ... set your cache to 2 days and you will end up with very little in Pending. I believe the stats (Sek, please help) say that on average over 90% are validated within 2 days. After that the curve slopes quite a bit so it does not help to set a larger cache, it only slows down other people from getting their validations ![]() |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I threw in the 1.5 days cache as a hint. With 2 days cache it might improve a bit, my worry being that if we all start caching like that, just because our personal desire is to not have PV, WCG ends up with millions of results stuck in IP, up front. So, cache if you need it, else just run with the flow, keep it to the default of 0.3 or whatever it presently is. Mind you, the single science exclusive WCG crunchers without the alternate work option might want a bit more... in case supplies runs low :D
----------------------------------------As for caching against general down time... touch wood, how often and how long does that take at WCG?
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
Movieman I confirm you have at the moment probably the most stellar machine if it crunches 24 threads at the time.
---------------------------------------- If I can do better in the future I will tell you. But for sure if we could use all the CUDA cores in a GTX200 series Nvidia card it would be nice. I have three of these cards available GTX285, GTX280, GTX260 on the machines but unused for WCG. A real pity. JM Boullier, yes you are right I take the numbers from My Result Status pages. So yes you are right I probably miss many of them. But how then should I proceed to be sure to have the true account ? ![]() ![]() |
||
|
|
|