Message boards :
Number crunching :
Bigger work buffer
Message board moderation
Author | Message |
---|---|
Mats Nilsson Send message Joined: 22 Oct 05 Posts: 26 Credit: 111,298 RAC: 0 |
I´m often getting thios message from projekt 2007-09-30 09:50:10|SETI@home|Requesting 3728 seconds of new work 2007-09-30 09:50:15|SETI@home|Scheduler RPC succeeded [server version 511] 2007-09-30 09:50:15|SETI@home|Deferring communication for 1 hr 28 min 38 sec 2007-09-30 09:50:15|SETI@home|Reason: no work from project My question is kan I get BOINC to download a bigger workbuffer for SETI but not for my other projekt. The reason is because I don´t have this problem with my other projekt and want to keep as low buffer as posible because I´m not a 24/7 cruncher and SETI has longer returntimes. |
Suzuki Send message Joined: 17 Sep 01 Posts: 318 Credit: 4,474,402 RAC: 1 |
You can change the number of days worth of work that BOINC will store for each project; I think they are separate for each project too. However, if you run out at the same time as Berkeley do, you'll still have to wait for the splitters to catch up and produce more new units. Cheers, Steve. |
Mats Nilsson Send message Joined: 22 Oct 05 Posts: 26 Credit: 111,298 RAC: 0 |
You can change the number of days worth of work that BOINC will store for each project; I think they are separate for each project too. I only find this under General preferances and it´s for all BOINC project´s I´m attached to.
Yes but if I have a bigger buffer Boinc will try to fill it up before it´s empty and giving me time to catch work before I´m empty. |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
AFAIK there is no way to specify that one project should get more buffer than another. After all, it's all timeshared with the CPU and if Rosetta needs more work but all you've got cached is SETI, then Rosetta is just going to have to download some anyway when your cache is full of SETI. This could have a negative impact on your shared time if you have some cached workunits that need to be crunched but you also owe some time to another project. |
DJStarfox Send message Joined: 23 May 01 Posts: 1066 Credit: 1,226,053 RAC: 2 |
I´m often getting thios message from projekt Just change the resource share of SETI to be higher than other project(s). It may take BOINC a few days of crunching to adapt, but it will download more WU. |
John Clark Send message Joined: 29 Sep 99 Posts: 16515 Credit: 4,418,829 RAC: 0 |
I am not sure whether this can be done. I've tried to reduce the share between S@H and Einstein towards S@H (currently 50:50). When I made the changes in General Preferences they were updated in the other project on the next contact. I think OzzFan is correct on this? It's good to be back amongst friends and colleagues |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I´m often getting thios message from projekt BOINC will take the resource shares, buffer size, and deadlines into account and keep a reasonable amount of work for each project. On the other hand, if you leave the buffer small, and you can't get work from SETI, BOINC will "loan" extra time to your other projects (to keep crunching) and give that time back to SETI when work is available. BOINC uses these "debts" to keep busy while also following your resource share. |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I´m often getting thios message from projekt The work fetch behavior is version dependent. 5.8 and earlier load a full queue from each project. 5.10 loads a fractional queue from each project based on resource shares. However, if one (or more) project(s) is(are) not providing work, other projects will get requests to fill the queue to at least minimum size. BOINC WIKI |
Uioped1 Send message Joined: 17 Sep 03 Posts: 50 Credit: 1,179,926 RAC: 0 |
Edit: I am summarizing and expanding on what others said before me... Edit #2: As was pointed out before, there are different behaviors for the different BOINC versions.
This will not help the problem you're experiencing. Unfortunately, as others have said, there is no direct way to increase your cache for SETI only. You have two options: If you do nothing, BOINC will remember how long you have been without SETI workunits, and the next time they become available, it will download extra work, up to the size of your work unit cache. [in boinc v5.10.20] If you find that this amount of work is still not enough, then you might consider the other option: Increase your overall cache size. You can feel secure doing this, because BOINC is smart enough to figure out that your machine is not a 24x7 cruncher, and will adjust the amount of work necessary [to fill your cache] based on the details of your particular situation. The only time where I wouldn't recommend this situation is if the amount of time your computer is on and crunching varies drastically, and over periods longer than your shortest deadline. In other words, if some days your computer is off, but some days it's on, that's fine as long as your deadlines are measured in weeks. If you sometimes don't turn the computer on for a week, keep a very short cache, and trust BOINC to get the work based on which project is 'owed' work the most. |
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0 |
The work fetch behavior is version dependent. 5.8 and earlier load a full queue from each project. 5.10 loads a fractional queue from each project based on resource shares. However, if one (or more) project(s) is(are) not providing work, other projects will get requests to fill the queue to at least minimum size. On 5.10.20 I've seen some oddities within this general picture. For example, during our recent SETI difficulty, one of my hosts kept unsuccessfully requesting small amounts of SETI even though it already had nine times its fractional queue share. Meanwhile Einstein had dropped to half its fractional queue share without generating any requests. This persisted for many hours and many dozens of SETI requests. However, with no intervention on my part, eventually a huge Einstein request was made, and filled, and the SETI requests stopped. But in general the behavior described above by John is a huge improvement in behavior under disturbed conditions such as home ISP outage or Berkeley server droughts for those of us who run some projects at low resource share. |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
The work fetch behavior is version dependent. 5.8 and earlier load a full queue from each project. 5.10 loads a fractional queue from each project based on resource shares. However, if one (or more) project(s) is(are) not providing work, other projects will get requests to fill the queue to at least minimum size. Of course there are other things that affect the work requests. For example if a project has had to borrow CPU time it will be prevented from downloading work for a time and its share of the cache will go to other projects. This is what Long Term Debt (LTD) ia about. If a project is not supplying work, its share goes to other projects. If a project is in deadline trouble, it will not have a work request made, and it is possible that some of its share of the queue will go to other projects. It is also possible that enough work will be in the queue so that it is full and a task is in deadline trouble. In this case no project will be asked for work. There are several causes for temporary share violations. BOINC WIKI |
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0 |
For example, during our recent SETI difficulty, one of my hosts kept unsuccessfully requesting small amounts of SETI even though it already had nine times its fractional queue share. Meanwhile Einstein had dropped to half its fractional queue share without generating any requests. This persisted for many hours and many dozens of SETI requests. However, with no intervention on my part, eventually a huge Einstein request was made, and filled, and the SETI requests stopped. Yes, but none of those situations applied in this case. |
Rowe Family and Friends Send message Joined: 25 Dec 00 Posts: 17 Credit: 38,395,231 RAC: 67 |
i have a simular problem: in bonic 5.10.20, and running seti i have set my additional work buffer to 10 day, and my connect every to 10 days. yet i only have about enough work for 2 days if that. every time i hit update it says not requesting new work my Laptop is finishing work units in about 3h and currently only has 13 units downloaded and qued up. id like to get a few more units downloaded as i tend to go away with work and dont have an internet connection for a few weeks at a time any help would be much apreciated |
Osiris30 Send message Joined: 19 Aug 07 Posts: 264 Credit: 41,917,631 RAC: 0 |
i have a simular problem: Sounds like your local preferences are overriding the website ones.. I'm sure someone here can talk you through fixing that. (I suck at tech support, sorry LOL) |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
i have a simular problem: Try setting "connect every" to something alot smaller, like maybe 0.5. BOINC gets nervous if you try to make it cache too much work. |
Rowe Family and Friends Send message Joined: 25 Dec 00 Posts: 17 Credit: 38,395,231 RAC: 67 |
i have a simular problem: i have no problems on my desktop getting new work when i set it to 10/10 just my laptop doesnt like it for some reason ive set it to 5/10 and its got 3 new units but thats still no where near enough for 10 days |
W-K 666 Send message Joined: 18 May 99 Posts: 19062 Credit: 40,757,560 RAC: 67 |
Assuming you have set the 10/10 preferences here on the Seti site in Your Account/General Preferences/location/Network Usage/, and they don't work, you need to check to see, on that computer if you have a global_prefs_override.xml file in the BOINC folder. If so then in the manager use the Advanced Menu/Preferences/Network Usage tab and then set your requirements there. As noted in previous posts, it doesn't like large connect intervals. You should also note that as the deadline for short units is eight days, once you get one of these, and there are quite a lot of them, if 'connect to network' is set over eight days then you will enter Earliest Deadline First (EDF) mode. EDF will stop all further downloads until these units are reported, as the BOINC manager will assume it is not going to contact Seti until the next 'connect to network' interval. Andy |
Rowe Family and Friends Send message Joined: 25 Dec 00 Posts: 17 Credit: 38,395,231 RAC: 67 |
Assuming you have set the 10/10 preferences here on the Seti site in Your Account/General Preferences/location/Network Usage/, and they don't work, you need to check to see, on that computer if you have a global_prefs_override.xml file in the BOINC folder. cheers for the reply yes i have that file on my laptop, and it has exactally the same info in it as my desktop has, and the desktop has no problem getting work for a 20/10 ratio, yet the laptop wont go more than 5/10 before it wont get anymore work. and even then it only has 16 units downloaded. 16 units and a 15day cashe just doesnt quiet add up, especially when it takes about 3h to process each work unit |
W-K 666 Send message Joined: 18 May 99 Posts: 19062 Credit: 40,757,560 RAC: 67 |
Have you looked at your computer stats like these for my pent M % of time BOINC client is running 99.9522 % While BOINC running, % of time work is allowed 99.8434 % Average CPU efficiency 0.962706 Result duration correction factor 0.273591 Any of the top three not close to 100% or a high RDCF will also lower downloads. Andy |
Rowe Family and Friends Send message Joined: 25 Dec 00 Posts: 17 Credit: 38,395,231 RAC: 67 |
% of time BOINC client is running 71.8875 % While BOINC running, % of time work is allowed 99.9883 % Average CPU efficiency 0.767899 Result duration correction factor 0.58026 the only reason why the top one is so low is because when im away with work and bonic finished all its units that are qued up i close it till i get back to an internet connection to upload them. surely this shouldnt be causeing me to only download about 3 days worth of units when its set at 10/10 |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.