Message boards :
Number crunching :
Not being impatient... just...
Message board moderation
Author | Message |
---|---|
Cath Send message Joined: 15 Dec 07 Posts: 16 Credit: 1,179,700 RAC: 0 |
So, I noticed that you can look at your pending tasks and see who your wingman is for them. I thought this would be an interesting way of estimating how long it would take to get my task reported based on the user's workload and RAC... Unfortunately, I noticed something that really seems inconsiderate. One of the tasks I have pending isn't being crunched at all. It's sitting in a task list that's around 400 tasks long (300+ of which are "new"). Granted, the user has a decent RAC, but this user has tasks not even started from the end of February, but is getting new tasks almost constantly. If it takes said user 2 1/2 hours to process each task and has a quad core computer, their time to complete all 300 of these is 187 hours plus... or around 7 days (of course, assuming that the user is only processing SETI tasks). I feel that requesting tasks 7 days before you will need them is a bit excessive and, as I said, inconsiderate to wingmen. Is there any other explanation that I'm missing in this case? Not that I think I'll be able to change anything about this but... *shrug* |
dnolan Send message Joined: 30 Aug 01 Posts: 1228 Credit: 47,779,411 RAC: 32 |
It's very possible that the user in question had an issue with Boinc and the older WUs are no longer on the system. Unless users in this situation detach and then re-attach, the tasks have to wait until they time out. You can usually tell this is the case because if you look through their tasks, there are some older pending WUs, then a bunch of status NEW tasks, then newer ones that are being returned. Another situation that can look similar is if they have a large cache, then get a bunch of "shorties" that have a short, 7 day deadline. Those go into high priority mode and crunch before older, non-shorties. You can usually tell this situation because it looks like the other one, but all the WUs being returned don't take very long to finish (compared to other completed tasks). -Dave |
MadMaC Send message Joined: 4 Apr 01 Posts: 201 Credit: 47,158,217 RAC: 0 |
I can see what you mean about this, but it is a personal preference. I run all my BOINC clients with a 10 day cache, as I like the idea of them just running and not having to worry about whether the server sending work has problems or not. That way I dont have to constantly check everything or worry about running out of work... I hope Im not annoying people by running it that way, I dont abort work units and get through them eventually - it all evens out in the end I s'pose |
Bernie Vine Send message Joined: 26 May 99 Posts: 9954 Credit: 103,452,613 RAC: 328 |
So, I noticed that you can look at your pending tasks and see who your wingman is for them. I thought this would be an interesting way of estimating how long it would take to get my task reported based on the user's workload and RAC... I actually agree 100% on this, I run 7 machines with a 1 day cache and one machine with a 2 day. So tasks are turned around quicker, if I have a major problem, i.e. HDD or MB failure, only a small number of WU's are delayed. If SETI@home has problems longer than 1 day I either switch my machines off(being green) or might crunch for WCG a bit. I would never dream of holding a 10 day cache seems selfish to me. But what do I know;-) Bernie. |
Borgholio Send message Joined: 2 Aug 99 Posts: 654 Credit: 18,623,738 RAC: 45 |
I tend to run a 10 day cache due to CUDA. It burns through Seti workunits so fast that I need to maintain a minimum of 30 - 40 per day to keep my GPU hot. I am aware of GPU grid but the performance issues with that project are worse than Seti with a VLAR workunit...so it's not really an option. You will be assimilated...bunghole! |
HAL Send message Joined: 28 Mar 03 Posts: 704 Credit: 870,617 RAC: 0 |
So, I noticed that you can look at your pending tasks and see who your wingman is for them. I thought this would be an interesting way of estimating how long it would take to get my task reported based on the user's workload and RAC... looks like you are learning Micromanagement techniques. This CAN be useful but time consuming. The object is to return YOUR cache ON TIME - HOWEVER if you MUST micromanage keep in mind it is YOUR cache that gets red marks - the other dude is on his own. I have wu,s assigned to dudes with 1000+ entries in their cache - so what, they responded a week ago and in MY cache they are suspended until SETI puts them on PRIORITY, in the meantime I process other client systems that are actively processing.Each week I have a small cadre of wu's that need to go active to clear MY Cache - usually totalling less than 18 hours. AP Wu's are another animal, 187 hours COULD me one AP WU - I had AP try to load on a 1GHz Pentium - 512 MB Ram and it came in at 500+ hours to process- needless to say It got dumped. The least stressful path is to let SETI manage the cache, but if you MUST micromanage - establish a routine that ALWAYS clears YOUR cache with no red marks, especially on those units where YOU are the wing. Classic WU= 7,237 Classic Hours= 42,079 |
Rick A. Sponholz Send message Joined: 13 Jun 99 Posts: 43 Credit: 23,179,807 RAC: 0 |
I'm not sure who your ISP is, but mine is AT&T and VERY unreliable. I can go days with no service at all, while I wait for network to come up, or for a repairman to show up. Now add Seti's well documented server problems and you get an idea why I keep a 10 day cach. You get your credit, just not instantly. |
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
I'm not sure who your ISP is, but mine is AT&T and VERY unreliable. That's strange. I've been with AT&T for years and they have always been VERY reliable. I've bever had a problem with them. |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
So, I noticed that you can look at your pending tasks and see who your wingman is for them. I thought this would be an interesting way of estimating how long it would take to get my task reported based on the user's workload and RAC... This is one of the effects of users being allowed to carry a larger-than-needed-cache of workunits. Each user is allowed to set up to 10 days cache plus a 10 day connect interval (meaning that the user doesn't expect to connect to the internet but every 10 days). One of the mindsets of many people is bigger, better, faster. In this case, a bigger cache means its better for them, without regard to their wingmen. Not that I'm advocating taking away this option, simply explaining the perils of doing so to your quorum partner(s). This is why I advocate a cache no larger than needed, such as my .333 connect interval with a 3.75 day cache, plus joining other projects just in case something major happens to the SETI servers to keep them down longer than my cache settings. |
W-K 666 Send message Joined: 18 May 99 Posts: 19065 Credit: 40,757,560 RAC: 67 |
Must admit that I don't understand the need for large caches of work. But would be interested to here from those that try to have a large cache. What is the largest average cache that can actually be achieved? If anybody would like to respond to this question, I would like the average to include a period when we have a storm of VHAR tasks. |
Cath Send message Joined: 15 Dec 07 Posts: 16 Credit: 1,179,700 RAC: 0 |
I appreciate all the responses. As I stated, I was mostly interested in seeing other responses, as well as a reasoning for why someone would do something like this. First off, I'm not worried about getting my credit. I just feel that it's a bit inefficient to have tasks sitting around growing mold. I might also point out that I'm a bit antsy for credit because I've just upgraded my computer and am finally getting tasks done and am excited to see the results in the graphics. I was so thrilled to see that I'd be getting over 1000 credits for each of the astropulse tasks I completed. My best day ever in the past is 712 credits, so... for all that I've been crunching for over a year, this is the first time I have a chance to make some decent headway. In some cases, I do understand the reasons for carrying a large cache. If SETI is, in fact, regularly on the fritz... I suppose it makes sense. If you have wonky internet, again, reason enough. Personally, I have cable internet, so I almost never have problems and, in the off chance I do, I can usually get it fixed within a day. I only run with a 1.25 day cache, which basically means one-two extra tasks, and I can get through the larger astropulse tasks in under a day, meaning whether I'm primary or wingman, it's done fast. HAL, I'm not sure I completely understand what you're saying. Perhaps part of it is that I'm not fully up on any lingo that might be used here regularly as I'm new to the forums. I'm not sure what you mean by "red marks"... if you're interested, of the 10 tasks I have pending, I'm wingman on 7... I'm sure this won't always be the case, but I, at least, hope to do my part to get tasks processed within a day or two of receipt. As you may have seen, I run two projects, so, if SETI is down, I'm sure Einstein will pick up the slack. ~C |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
I see you are running the optimized client for Multi-Beam, but do not have the optimized AP client yet. http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=176 That should speed up AP a little bit. |
ccappel Send message Joined: 27 Jan 00 Posts: 362 Credit: 1,516,412 RAC: 0 |
http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=176 A lot bit. |
Cath Send message Joined: 15 Dec 07 Posts: 16 Credit: 1,179,700 RAC: 0 |
I see you are running the optimized client for Multi-Beam, but do not have the optimized AP client yet. Do I do the same thing as with the optimized client? Copy to the library, change the owner name thingy...? Edit: So, I looked at the files that came in that download. They're the same as what I've got already. The task output for my most recent results says that I'm running astropulse, so I think it's all fully installed. |
Rick A. Sponholz Send message Joined: 13 Jun 99 Posts: 43 Credit: 23,179,807 RAC: 0 |
I have a fiber optic DSL line with AT&T, and when it's up, it works very well. I keep BOINC connected 24/7, so my completed work units upload as soon as they're completed (as long as AT&T is working) Rick |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
|
Cath Send message Joined: 15 Dec 07 Posts: 16 Credit: 1,179,700 RAC: 0 |
I'm not trying to quibble and I appreciate that you're trying to help me get the most out of my crunching but if you look at a most recent non-astropulse task, it looks exactly like your results on your Mac. When I open the file you linked me to, app_info.xml, in text edit it is a shorter version of what I have already in the app_info.xml file you gave me a few days ago. Additionally, since they both have the same name, I would have problems actually using the program. The other two items are the two exe files, which, again, I have already (confirmed using md5 hash to check the sum of the file). If it's not using the optimizer to crunch, it's not because I don't have it. If this is the case, it's possible that I did something wrong when installing the optimizer. Currently, it's in HD>Library>Application\ Support>BOINC\ Data>Projects>setiathome.berkeley.edu and the owners are set to boinc_master as I was told to. I understand that running the installer would ensure that I've done everything correctly but, considering that at least part of it's running properly... Would it be best to just re-install Boinc? |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
There are a number of good reasons to keep a large cache, and there are a number of relatively poor reasons (mostly paranoia) for keeping a large cache. Picture someone who takes their laptop on vacation, with no internet access, they aren't reporting for a week. Another member was allowed to let his office computers crunch, but only contact SETI during set hours. It is just part of the game. I wouldn't characterize it as inconsiderate. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Must admit that I don't understand the need for large caches of work. But would be interested to here from those that try to have a large cache. ... something around half the shortest deadline, or about two days. I'm assuming worst-case everything. There is still benefit from longer caches, but BOINC may not keep the cache completely full when we have nothing but shorties, and it may not stay completely full when the DCF is not exactly right due the differences in processing time vs. angle range. There are fewer issues with longer deadlines and more consistent run-times. That makes predicting the future easier, at the expense of less efficient crunching. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
... something around half the shortest deadline, or about two days. The shortest deadline on SETI is 7 days, so half that would allow ~3.5 days. We used to have 4-day deadline work with the old linefeed antenna application, but not since the transition to the multibeam recorder and the matching application. |
©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.