Message boards :
Number crunching :
Up and Running Again
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
Send message Joined: 24 Oct 04 Posts: 1130 Credit: 49,848,934 RAC: 8,513 |
Server Status Up, 122,798 workunits to crunch 63434 workunits in progress 39 concurrent connections I had the power out here from a wind storm. last weekend until monday night and I was wondering what I would find left here. was surprised to see lots of work still here then every time I check the number goes up instead of down so maybe I didn't lose too much time here. Back to work............. Volunteer Mad Scientist For Life |
Send message Joined: 24 Oct 04 Posts: 1130 Credit: 49,848,934 RAC: 8,513 |
Server Status Up, 9,222 workunits to crunch (8,645 ten minutes later) 65510 workunits in progress 42 concurrent connections Running low........and I am loaded and ready Volunteer Mad Scientist For Life |
Send message Joined: 13 Nov 05 Posts: 1 Credit: 3,388 RAC: 0 |
Hum one of my hosts with about 10 WU buffered suffered a HDD failure... it will probably not got online again (it was pretty old). Is there any way for me to force WU reassignation? Sorry if this post doesn't belong here. Cheers |
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
Hum one of my hosts with about 10 WU buffered suffered a HDD failure... it will probably not got online again (it was pretty old). Is there any way for me to force WU reassignation? No practicable way. This project is designed on the assumption that things like this will happen: 5 copies of each of those wu will have been sent out, and as soon as 3 come back that agree it won't matter about your results not being returned. If it happens that the project can't get 3 results to agree from the other 4, a sixth copy of that WU will be sent out soon after the deadline expires. It is a matter of boinc policy that there is no web-based method for cancelling results once issued; if I understand this correctly it is felt that such a feature would generate user errors and tempt some people to abuse the deadline system and that these issues would outweigh the advantages. Thanks for the thought, but you can sit back and let the system catch up with the problem in its own time. Please note, this is very different from people abusing the deadline system - delays such as yours are inevitable and in a big project like this can be predicted on a statistical basis. These anticipated delays are built into the timings and the project team can respond to the stats by adjusting the number of results sent out for each wu. You may have seen posts objecting to people who deliberately take more than they can crunch. That is different because it adds an unpredictable unnecessary and additional delay to the process. Critics of the practice say also that it is greedy precisely because it is deliberate.
Personally I'd have started a new thread for your question, (on this same Number Crunching board), as I like to limit 'topical drift'. It is not clear-cut: conversations do naturally drift, it is hard to know when to break the flow. River~~ |
Send message Joined: 24 Oct 04 Posts: 1130 Credit: 49,848,934 RAC: 8,513 |
That's ok River......somebody must read these and get a tip now and then Last year I tried to get my old PIII 500 to run a few LHC's and soon after it had a motherboard problem so it was unable to ever finish and send them in. I put the HD in another machine and dumped them back right away and retired that pc for good. (the pc I put that HD in to do that was even slower and never would have finished any in the PII) And I only load my faster pc's up with what they can handle (the server helps with that too as you know) One of mine finished already and the other 2 are close to being finished so they will be ready for more soon and just running Einsteins if we still have to wait for more LHC's........and I don't run any of the other projects myself. Volunteer Mad Scientist For Life |
Send message Joined: 30 Jan 06 Posts: 10 Credit: 1,947 RAC: 0 |
I wonder if in order to avoid these outages while they are generating new work if it would be within their goals to prepare 2 simultaneous experimental paths (for lack of a better term) so that the 2nd LHC 'batch' is ready to load while they compile and redirect the first batch. LHC goal '2' doing something useful so that the computing power available out here in hostland is not neglected. Just a thought. Founder of BOINC group Objectivists: Rational people crunching data for science. |
Send message Joined: 1 Sep 04 Posts: 275 Credit: 2,652,452 RAC: 0 |
I wonder if in order to avoid these outages while they are generating new work if it would be within their goals to prepare 2 simultaneous experimental paths (for lack of a better term) so that the 2nd LHC 'batch' is ready to load while they compile and redirect the first batch. LHC goal '2' doing something useful so that the computing power available out here in hostland is not neglected. Just a thought. I believe this is already happening. The reason I think this is there has been at least one time when new work appeared before we actually ran out and also there have been very short periods of no work. I would think if there was only one path of studies the period of no work would be more consistant and there would never be new work without a period of no work. BOINC WIKI BOINCing since 2002/12/8 |
©2024 CERN