| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Retired Forums
Forum: Member-to-Member Support [Read Only]
Thread: Grid slowing down?! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 14
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
This is a graph of daily results posted v 7 day moving average.
 Things are slowing down! Keep your computers on - get more friends involved and to stay involved. Lets get that line pointing upwards again. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Slowing down?.....The other grid has come to a complete halt
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
My guess is that vacation and warmer weather are taking their toll. I know that it's been very warm in my house, so I've been pausing the agent more often to try to cool things down (not that it actually works... but it's one of those things that makes sense in your mind).
You can also see that the results are very cyclical... Maybe with longer work units, it's taking people longer to complete results. From the graph it looks like we've had drops before and rebounded from those (espeically right after 141). I hope we can do the same now. |
||
|
|
RT
Master Cruncher USA - Texas - DFW Joined: Dec 22, 2004 Post Count: 2636 Status: Offline Project Badges: 
|
While I do not know any more about it than you guys apparently do, my assumption is that this is just the big work units introducing these statistical perturbations. I suspect, and assume that this is just the natural valley and not a reduction in overall CPU time or points being generated (though the reporting of these is time shifted a bit).
----------------------------------------These other two measurements are not as perturbed by the WU size changes by their very nature. When smaller work units come around again (assuming they will), then we should see the mirror effect. All this is just an assumption on my part, but it sounds good to me.  And could be argued by looking at the graph. Have a nice day everyone! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ah, my favorite topic!
I think there are a lot of churn in users who are actually returning results underneath the high level statistics. The most obvious one to me is the fact that while number of users has increased steadily from the late January, the number of points has remained essentially flat - i.e., ranging between 10,000,000 and 15,000,000. FWIW, looking at the number of results can be misleading as the size of the work units can vary. Most direct measure is number of CPU hours reported. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If you look at http://vspx27.stanford.edu/DCcomparison.html you can see the number of CPU hours divided by 24, which gives the number of equivalent full-time computers. We are down to about 19,000 computers. For a while, we were over 21,000. A similar site used to include the BOINC projects, until a change in statistics reported for BOINC broke it. We seem to slowly lose computers until some publicity event brings in a flurry of new members.
|
||
|
|
Viktors
Former World Community Grid Tech Joined: Sep 20, 2004 Post Count: 653 Status: Offline Project Badges: 
|
Yes, just looking at results count can be misleading in our project because work units vary quite a bit in run time, and run time is not actually cpu time. For example, there is a particularly difficult batch of proteins being worked on right now, so the work units are averaging longer run times than normal. Also, with the onset of warmer weather in the northern hemisphere, I wonder how many people don't run their heat generating PC's as many hours.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
fj wrote
I think there are a lot of churn in users and I think this is where we should look for the problem. The WU size averages out over all the users, so it will affect the number of results returned however should not (on average) effect the point generated stat. (except for the first day a huge batch of large WU is distributed) Churn is always a problem when dealing with services that require an effort to do something over time. I am sure we could come up with a lot of ideas on how to deal with the churn, but the real question to be answered is "who is generating the churn?" Are we loosing older members who are getting bored or new members that come and go. This is critical to know as the actions taken for these two segments will be very different. If there are any supporters of this approach, then lets brainstorm it and comeup with one or two ideas to implement and test. This will help us resolve the speculation. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
great job changetheworld
The trend line on that graph goes up and now it is dropping. This situation needs to be addressed. Do you have similar data for the average points per day - this should give us a better indication of the computing power trend for the project. Thnks |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
We're watching the graph too. This is our first summer to run the grid and the rate of crunching perhaps does drop during this season. Another thing might be that the Human Proteome Folding project is in the doldrums. At 60 to 67 percent complete, it still has a way to go.
Participation might also be a function of media coverage. Later on this year when the Linux client comes out, the next research project gets added, and the Human Proteome Folding scientists start publishing results, then there'll be plenty of media to go around. Right now, the only excitement is in the forums. |
||
|
|
|