Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Completed Research Forum: The Clean Energy Project - Phase 2 Forum Thread: I note run-time per CEP2 task is showing a decided trend upwards...[RESOLVED] |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 9
|
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
...both through personal observation and from looking at sekerob's most excellent chart . Is this an anticipated trend, and is it expected to continue?
----------------------------------------[Edit 1 times, last edit by Former Member at Apr 24, 2012 11:14:38 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I thought they said they distibuted x amount of those small tasks, as Sek noted if you look at the C in the beginning part of the WU, they higher the number the longer it takes.
----------------------------------------Thought they dipped down to a batch of small WU C22-23? maybe less, and it seems we have been working our way back up, and I would personally expect it to continue. EDIT: found cleanenergy's response from short WU: Hi everybody we mentioned this in another thread already but just to make sure that no one is left in the dark: We have submitted a bunch of smaller molecules to the feeder a while ago and these will run quite quickly. There are about 2-3 million of these smaller motifs in the queue until we reach previous levels. The size will gradually increase as it has done before. Hope this explains the recent behavior - no need to worry at all! Best wishes Your Harvard CEP team [Edit 1 times, last edit by Former Member at Apr 21, 2012 5:24:38 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I thought they said they distibuted x amount of those small task I believe that if you look at the date of that post from the CEP team, it was February 22nd...which corresponds to the short-lived but significant dip in the run-times for CEP2 that you see in sekerob's linked graph.[...] EDIT: found cleanenergy's response from short WU: Hi everybody we mentioned this in another thread already but just to make sure that no one is left in the dark: We have submitted a bunch of smaller molecules to the feeder a while ago [...] Your Harvard CEP team The trend I am referring to is the current average of approximately 8.1 hours, which exceeds by an hour/is a 14% increase in the previous high of approximately 7.1 hours on the same graph. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Good observation, ibsteve2u.
My guess is that we are moving through a library of molecules that are organized somehow into similar shapes. So sections of the library will require similar computation times that increase or shorten in unison. Just a guess on my part. Lawrence |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi everybody,
5pot found the relevant old post and I guess that says it all. Best wishes from Your Harvard CEP team |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ah, ibsteve, now I see what you're saying. Cleanenergy, the question is about "how high" will these run times get in the long run? After that batch, we picked up where we left off, and has been steadily increasing. How much "bigger' will the runtimes get? Or, maybe how much bigger will the molecules get etc.?
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There is no fixed limit on how big these can get. We currently run the very big ones on high-performance clusters and will keep digging away on the smaller ones on the grid. Once this starts becoming unsustainable, i.e., in terms of RAM and because we don't get through a sufficient part of the workunits anymore, we will change to a different workunit type. But we don't expect this to happen any time soon.
Best wishes Your Harvard CEP team |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The "most excellent chart" is apparently not good enough since somehow 8.1 hours mean was read from the Y-Axis, when in fact the data and table on right for current day and 7 day average indicates 7.6 hours until yesterday... so have granulated the axis further to show 20 steps stead of 10. Have also overlaid a centered mean trend line on top of the daily curve. Obviously, being centered, the curve will lack 4 days.
Just today we have a little spike of 7.9 hours in morning session. Also, whilst the chart goes back 6 months, Aug-Sept 2011 we ran close to 9 hours average on a near continuous basis with a peak day of 9.1 hours, so don't see anything extraordinary... the repetition of a trend while working through a regular structured compound libraries. Whence we get to have ''no time limit" [optional for the volunteer as well, which I think to have seen will still be constraint at 24 hours], we'll learn if this has any impact upwards [and maybe slower device owners bowing out]. Only seeing 1 in 20 on my near 6 year old quad that actually hits the 12 hour barrier. --//-- |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Only seeing 1 in 20 on my near 6 year old quad that actually hits the 12 hour barrier. Me, I had a 12 hour on an i7-950 yesterday...in fact, 50% of my valid results as stored at WCG (only most of April's retained) are > 8 hours across systems including the aforementioned i7-950, an i7-2600K, an i7-980x, and a (the slow one, relatively speaking...although it isn't trending higher than any of the other ones, and its max run-times are lower, in fact) Q9550.--//-- No biggie. Thought I'd ask about the times to see if there was a predicted/predictable trend. Doesn't appear that there is, so if I don't have an opportunity to plan for increasing run-times and they begin cramping other things that I'm doing, I'll shut 'em down...problem solved. |
||
|
|