Message boards :
Number crunching :
Panic Mode On (24) Server problems
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 11 · Next
Author | Message |
---|---|
Francesco Forti Send message Joined: 24 May 00 Posts: 334 Credit: 204,421,005 RAC: 15 |
Personally, I don't like the idea of getting shot down by friendly fire. ;-) Yes, your are right, but 42 days of deadline as I see - are too many. How many results, not reported in the 30-35 days, are reported in 42 days? I don't know but admins do, I think. Better to have shorter deadlines (ok, without any retroactivity) and reasign task asap. It's not my problem. It is the space on disk of the project. Bye FF |
Blurf Send message Joined: 2 Sep 06 Posts: 8962 Credit: 12,678,685 RAC: 0 |
|
Vistro Send message Joined: 6 Aug 08 Posts: 233 Credit: 316,549 RAC: 0 |
A law that specifically prevents money to go to SETI... iiiimmmmmmmmmmm a dumbass.... lol. BOINC is being a little wonky. I'm trying to cut down on how often I use the network, I monitor the different graphs and enable network communication when it's low, but BOINC wants to do it other times. What I do, is to help the Replica DB catch up, is I let the tasks upload, and then report/update twice a week. Well, after all of the uploads, it tries to update, even if I keep the network comms disabled the entire time. 30+ Computers heading our way! Currently at the "Zomg we need to talk to our tech expert at the co-op about this first!!!" stage. 16 Lab machines and 14+ Staff machines each with 2.2Ghz CPUs and 256MB ram. Think they balance? The RAM certainly is bad |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Personally, I don't like the idea of getting shot down by friendly fire. ;-) Before we can even talk about what the deadline should be, we need to talk about why we even have deadlines. In a perfect world, deadlines are not needed. Work would go out to be crunched, get crunched, and would always be returned. We do not live in a perfect world. Computers die. People go on vacation and turn them off without thinking (okay, no one who visits the forums, we're exceptional), they quit without giving a second thought. Lot of reasons. Deadlines exist to detect lost work. So, the next question is: what happens if the deadline is too short? A machine crunches constantly and diligently, and returns the result late. In the meantime, work is reassigned, and the third cruncher does work that is not strictly needed -- it could have done a different work unit, or if SETI is short of work, it could have done work for another project (or we could have saved the electricity, or who knows what). A side effect of short deadlines: you eliminate those with slower machines from participating, except on the shortest deadlines. The opposite question is: what happens if the deadline is too long? Because work is not reissued until the deadline, lost work takes a little longer to detect. It takes a little longer for that work unit to validate. If there is enough work, and every work unit is equally likely to produce a result, then reissues are not more important than original work units. Yes, they do have to be kept in the SETI database, but if you're worried about the overall database size, just purge completed results as soon as they validate. I don't think we should be changing deadlines to "help out" the database. Better to put that effort into data recorder problems. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Agreed, but just to elaborate on your comments about the deadlines a bit. It's not so much how long the deadline is in absolute terms, but what I like to refer to as the 'Tightness Factor' of the deadline. IOWs the ratio of how long the task will actually take compared to the time allowed to run by the deadline. IIRC, with the last deadline settings made, shorties are actually a little higher TF than the longer deadline tasks in MB. <edit> Correct that, it's the other way around. The longer running AR's got their deadlines shortened up a bit, while the shorties stayed pretty much the same. However, when the resolution of the analysis got doubled recently, a side effect was the TF also doubled since the deadlines stayed the same. Another effect of a high TF is that it makes a project less 'friendly' in multiple project scenarios, even if you run modest cache and Resource Share settings. Anyone who has run the ultra tight deadline tasks at MW knows all about that! ;-) Alinator |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Naw, you're just a little impatient. Getting the whole picture takes time, and a bit of work. We all shoot from the hip sometime.
This is as designed. If BOINC can talk to the network, it's going to do everything it can. Reporting is good -- and the replica is current last I checked. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
I let Boinc do its thing. When a work unit is done i let boinc call it in. To save up a bunch of work units to report could run you into problems Of geting rid of them. I hated the no upload stuff we have gone through not to long ago. Me Id say get them back homne as fast as boinc will take them. And i will be shutting the i7 down shortly. ran out of AP and seeing as I have to work tonight, no use leaving this computer on. Another computer getting a well earned rest. [/quote] Old James |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Might as well take advantage of the downtime and do maintenence on them while they're resting. Then they'll be ready for a full scale assault on getting your RAC back up to where it was. ;-) Alinator |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
... Deadlines doubled along with estimates and work, hence 'Tightness Factor' wasn't affected by that change. One very minor exception was that prior to the change Very High Angle Range 'shorties' had an exact 7 day deadline because the splitter has that as a floor, after the change 'shorties' have the calculated 13d18h16m40s. Joe |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
... My bad, and I stand corrected. ;-) I stopped my in depth data logging at the beginning of the summer, and only gave a cursory look at the deadlines online when the change was made. Must have been looking at the wrong tasks when I drew that conclusion. :-O Alinator |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
Might as well take advantage of the downtime and do maintenence on them while they're resting. That is so true. Blow out the old cobwebs and stuff:) But i have also just set NNT and cleaned out my fans evry now and then anyway. [/quote] Old James |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Agreed, but just to elaborate on your comments about the deadlines a bit. All true. I think it's a mix of "tightness factor" and maybe "dedication factor" -- you need to take the tightness factor, which is relative to performance, and the "dedication factor" which is a whole mess of things like what happens if the owner goes on vacation, do they shut down on Friday night and start over on Monday morning, etc. But the other part of my comment is that a deadline that is a minute too short has a bigger impact than a deadline that is way too long. ... especially when we're talking about catching exceptions. |
Vistro Send message Joined: 6 Aug 08 Posts: 233 Credit: 316,549 RAC: 0 |
If the resolution was doubled, does that mean we have to do the old ones over again? 30+ Computers heading our way! Currently at the "Zomg we need to talk to our tech expert at the co-op about this first!!!" stage. 16 Lab machines and 14+ Staff machines each with 2.2Ghz CPUs and 256MB ram. Think they balance? The RAM certainly is bad |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
In theory, there might be some benefit to rerunning the old data whith the new apps or split settings. Back when he original Enhanced app and then MB first came out they did just that. However, I don't think they've have considered that option for a while, except for 'tapes' which failed to completely split succcessfully for one reason or another. That might be changing, if the fresh data sitch from Arecibo doesn't improve some. Alinator |
Cappy [Team Musketeers] Send message Joined: 19 Feb 03 Posts: 18 Credit: 1,180,143 RAC: 0 |
|
Ianab Send message Joined: 11 Jun 08 Posts: 732 Credit: 20,635,586 RAC: 5 |
nothing uploading or reporting,,,, its 4:56 EST Internet link is running at 95%, flat out. So the new work must have arrived, new work units for all, as long as you can get them. But uploads will probably be a bit patchy for a day or so. Ian |
rob smith Send message Joined: 7 Mar 03 Posts: 22227 Credit: 416,307,556 RAC: 380 |
For "a bit patchy" read "Totally stalled". I've been getting the same error message (give or take the times and back-offs) 09/09/2009 10:45:44 SETI@home Started upload of "wu id" 09/09/2009 10:46:54 SETI@home Temporarily failed upload of 31jl09ad.4673.3339.16.10.53_1_0: transient upload error 09/09/2009 10:46:54 SETI@home Backing off 2 hr 10 min 19 sec on upload of 31jl09ad.4673.3339.16.10.53_1_0 since 8amBST. I've more results to dispatch now than I had then... Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
HAL Send message Joined: 28 Mar 03 Posts: 704 Credit: 870,617 RAC: 0 |
More like Catch22 for days - no work Now - work available BUT ya can;t get it 'cause ya can't upload. |
Dave Send message Joined: 29 Mar 02 Posts: 778 Credit: 25,001,396 RAC: 0 |
Patience everybody... |
RandyC Send message Joined: 20 Oct 99 Posts: 714 Credit: 1,704,345 RAC: 0 |
More like Catch22 I got some! GPUs no longer idle...for now. If the uploads stay stalled I won't be able to download more... |
©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.