| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: More details needed |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 11
|
|
| Author |
|
|
64bit
Cruncher Joined: Apr 30, 2007 Post Count: 7 Status: Offline Project Badges: 
|
Hi, I'm new here and would like to suggest these information to be added either at WCG webpage or posted as sticky in the revelant WCG forum sections.
1. Due time for each project : I can't seam to find how long am I allowed to run each project as my slow PC is not run 24/7. 2. An approximation time needed for each project : How long does it takes to complete a work unit? Maybe have a 1Ghz PC for a guide. 3. Details of multicore support in BOINC : How does BOINC support multicore? Does it process a single work unit by utilizing multicore or does it downloads and process 2 or more work units simultaneously? Hopefully these information can be added and be updated to allow users to have better management on which projects they can participate and to prevent expired work units. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello 64bit,
1. Due time for each project : I can't seam to find how long am I allowed to run each project as my slow PC is not run 24/7. This can change rapidly for each project. The last notice I received was a few months ago when it was 1 week for one project, but recently I read a post in which a member said that it was 9 days for all projects. I just checked my BOINC Results Status page and my current HPF2 work unit is due back in 9 days. 2. An approximation time needed for each project : How long does it takes to complete a work unit? Maybe have a 1Ghz PC for a guide. I have suggested adding this to the minimum requirements page, though it is also quite changeable. You can get an idea about relative times by going to Statistics - By Projects - [project name] and looking at Run Time Per Result. 10 hours and 17 minutes for Fight AIDS At Home shows that my computer is faster than average since I do them in under 8 hours. Genome Comparison only takes a third as long, HPF2 about 2/3 as much as FAAH, HCMD is 2 hours longer than FAAH. 3. Details of multicore support in BOINC : How does BOINC support multicore? Does it process a single work unit by utilizing multicore or does it downloads and process 2 or more work units simultaneously? No sensible DC project would use multi-core processors by running multiple threads from a single work unit. The extra programming would not result in any greater throughput. That sort of technique is only used to get a single job done faster using a given number of computers. It does not increase the amount of work that can be done by those computers in any given time unit. BOINC simply runs more jobs if it finds more cores (if the user gives it permission to do so). All three of your ideas are good. Uhh. . . we don't have stickies. We are still using the initial 2004 version of the forum software and stickies were added in the 2005 release. Someday . . . when we have just long running projects going on and no new projects expected for a while . . . we will take on the task of updating the forum software. The best way would be to add this sort of improvement to the website. Lawrence |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Hello 64bit, 1. Due time for each project : I can't seam to find how long am I allowed to run each project as my slow PC is not run 24/7. This can change rapidly for each project. The last notice I received was a few months ago when it was 1 week for one project, but recently I read a post in which a member said that it was 9 days for all projects. I just checked my BOINC Results Status page and my current HPF2 work unit is due back in 9 days. Here's a post by 'a member' with some more elaboration on the subject: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=12953#94860 The technicians have recently been experimenting with the BOINC scheduling / due dates as the latest recommended version has a strange behavior (the bigger the buffer setting, the less work is actually given!!!!). Best max is about 2.5-3 days for pre-fetching work for storage in the local queue. A future release is scheduled to have much better control and buffering, so it might go back to the 7 days it was from mid 2006 to about early March 2007. Generally > 90% of the work is actually returned within 2-4 days to complete the quorum allowing x-validation and computation / assignment of credit (points).
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
No sensible DC project would use multi-core processors by running multiple threads from a single work unit. Hmm, don't think Folding@Home would agree with you...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ouch! Good point, Ingleside.
Some of the F@H programs try to reach a certain point as fast as possible using some random variations, then it is done and the server sends out the next step. Of course, all the work units that did not reach that point are unnecessary. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Multithreading is good, and can make applications massively parallel. But, when doing heavy duty crunching, there is no point having multiple threads on a single core. It merely creates cache contention and other resource contention. Knreed could explain it much better (and probably has in the past, if you care to search).
The other consideration is the origins of the science programs. Most were written to run on supercomputers. This is why all the current WCG projects take advantage of the BOINC 1-core, 1-process setup, and don't attempt multithreading. F@H have the luxury of actively developing their application. WCG mostly just enable existing applications, and "board" them to the grid. However, if core counts increase, and memory demands increase, then multithreading is the way to go (although it would still be 1-core, 1-thread, most likely). |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
The BOINC client right now assumes that 1 science app running will roughly use 1 processor. Thus it will start X number of workunits where X = number of processors on the computer (or the limit you set in your profile). Thus if a science application runs such that it uses more then its one core it will start to compete for cpu time with the other workunits that are running.
----------------------------------------BOINC is looking at changing this so that a workunit can specify that it will use X processors. This will allow the developers of the science applications feel free to use multi-threading if it would be of benefit to their application without having to be concerned about the impact on other workunits running. Additionally, as Didactylos noted, the science organizations that we collaborate with provide the applications. We work with them to make them suitable to run in the grid and do the work to 'board' them but we do not make fundamental changes to the application. In case you are curious, here is a list of the number of computers by number of processors for those that have communicated with our servers at least once over the past week: 21851 1 [Edit 1 times, last edit by knreed at May 2, 2007 4:03:05 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Interesting, as these machines have a total of 69415 cores and WCG members produced on Monday 62,731 CPU days equivalent of results returned. That's gives 90% CPU utilization were they all talking to the servers on a single day.
----------------------------------------W00T
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
I find this number of computers reporting during one week (43516) small versus the total number (about 280000 before the transition). Is it a count of BOINC machines only or does it include all the UD machines too?
----------------------------------------Jean. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi JmBoullier,
The UD client does not return any information about the number of processors. After all, it only uses one. So the numbers that knreed posts must be for BOINC. A recent upgrade to BOINC started it returning some very detailed information so that projects could be compiled to take advantage of the capabilities of the computer. This is a very new feature and none of our projects are compiled to take advantage of it. Lawrence |
||
|
|
|