| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: A 3 part Question on BOINC? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 34
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
A. Is there anywhere to be seen how much BOINC is contributing to the project v.v. UD agent
----------------------------------------B. Then how many pure net results are being returned (can it be said to simply take total BOINC returns divied by 4, since seemingly always 4 results are requested, though quorum of 3 determines the points). C. Finally, whilst under UD, the ratio hours (clock) to points seemed fairly steady, under BOINC the hours (real CPU allocated time) seems to wildly vary. I considered that FA@H is within fairly straight, but Rosetta is truelly all over the place.
WCG
Please help to make the Forums an enjoyable experience for All! |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
A. Is there anywhere to be seen how much BOINC is contributing to the project v.v. UD agent BOINC claims significantly more points than UD. When I glance through individual items on my Results Status page, I notice that about 1 of every 4 point claims for each result are much higher than the other 3. I assume, perhaps incorrectly, that the high claims are from BOINC. That suggests to me that BOINC handles around 25% of the work but my "analysis" is rather shoddy and may be very inaccurate. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The results status page only shows BOINC results. The UD system is seperate, and runs with a different batch of work units.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The results status page only shows BOINC results. The UD system is seperate, and runs with a different batch of work units. Ok, I believe you but darn, that blows a hole clean through another theory I had accepted as fact. Now I have a question but I will pose that question in a separate thread as it is off topic here. |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
A. Is there anywhere to be seen how much BOINC is contributing to the project v.v. UD agent Sorry - we haven't had time to add this information. I can tell you that BOINC agents contribute about 14% of daily run time (UD agents contributes the other 86%) B. Then how many pure net results are being returned (can it be said to simply take total BOINC returns divied by 4, since seemingly always 4 results are requested, though quorum of 3 determines the points). For BOINC on average, we send out 4.2 results per workunit. We send out 4 results for each workunit initially and then we send out additional results when errors are returned or when a certain amount of time passes. We get about 3.6 results back for those 4.2 results that we send. The difference between the 4.2 and 3.6 is largely due to results that are never returned. However a small proportion is also to due to various forms of errors. We are looking at ways to reduce these numbers becuase this would improve our efficiency. We cannot reduce the validation requirement of 3 workunits becuase we need this to ensure that the data is reliable. However, we are looking to see if we can reduce the number of results that we send initially from four to 3 while still being able to validate the workunit within a reasonable timeframe. We implemented a change a couple of weeks ago that will send any additional work for a given workunit (after the initial work sent out) to a 'reliable host' rather then any random host. This is looking promising but we want to monitor the effects for a bit longer before making any changes to the number of results sent. [Edit 1 times, last edit by knreed at May 19, 2006 2:55:43 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
We implemented a change a couple of weeks ago that will send any additional work for a given workunit (after the initial work sent out) to a 'reliable host' rather then any random host. Wow, I think I mentioned the possibility of doing that a few weeks ago. Did you use my idea, or was it in the works already? Either way, it's good to know the idea was a useful one. |
||
|
|
depriens
Senior Cruncher The Netherlands Joined: Jul 29, 2005 Post Count: 350 Status: Offline Project Badges: 
|
Good to hear that you guys are constantly improving "our" grid!
---------------------------------------- Thanks a lot for keeping us addicted to this!  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
We implemented a change a couple of weeks ago that will send any additional work for a given workunit (after the initial work sent out) to a 'reliable host' rather then any random host. This is looking promising but we want to monitor the effects for a bit longer before making any changes to the number of results sent. Good work, knreed!! Glad to see you're implementing these changes slowly and cautiously. Thanks for the other info too. Sending the 4th WU to a "reliable host" rather than a random host is an excellent idea. But going one step further... wouldn't it be nice if WUs could be flagged as either high priority or normal priority. BOINC and UD could spot the high priority flag and move that WU to the top of the host's work queue for that project or even to a position ahead of all projects if deadlines won't be missed. I am aware that UD doesn't normally cache WUs on the host but I think UDMon or some other addon does cache for UD. So maybe UDMon would have to be able to spot the flag too. Just some thoughts for future revisions. |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges:
|
Workunits are now being distributed that only send out three results at the start. We will be evaluating the effectiveness over the next couple of weeks. I will let everyone know what the new send/recieve ratio per workunit is once we have sufficient data.
----------------------------------------From my post above, the current average ratio is that for each workunit 4.2 results are sent and 3.6 results are received. [Edit 1 times, last edit by knreed at May 19, 2006 2:59:29 PM] |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges:
|
Wow, I think I mentioned the possibility of doing that a few weeks ago. Did you use my idea, or was it in the works already? This idea was originally batted around between Dr Anderson and the head of the Einstein@Home project back in Dec. During a conversation earlier this year Dr. Anderson mentioned it to us in conjunction with some of the other changes we gave back to BOINC. I went ahead and implemented it a few weeks ago. Once we have proven its usefulness we will give the code back to BOINC. [Edit 1 times, last edit by knreed at May 19, 2006 3:03:37 PM] |
||
|
|
|