Ghost Crunchers / computers
Message boards :
Number crunching :
Ghost Crunchers / computers
Message board moderation
Author | Message |
---|---|
Send message Joined: 10 Jul 13 Posts: 21 Credit: 10,363,908 RAC: 5 |
Last modified: 17 Apr 2017, 6:32:57 UTC Hello, Would it be possible to prevent "Ghost Crunchers / computers" to upload a full cache of WU's (sometimes > 1000) without calculating them. The server already cancelled some of these uncalculated WU's, will cancel the remaining WU's automaticaly, but these crunchers will again and again receive a brand new batch of WU's that they won't return in time. As I was wondering why I have so many "waiting validation" (> 500), I found at least 3 or 4 computers having > 1000 WU's + 0 returned WU since April 11th and already several series of "cancelled by the server" WU's. Thank You Best Regards |
Send message Joined: 10 Jul 13 Posts: 21 Credit: 10,363,908 RAC: 5 |
|
Send message Joined: 28 Sep 13 Posts: 29 Credit: 120,331,090 RAC: 14,573 |
Last modified: 26 Apr 2017, 3:43:22 UTC Kyong... Read and laugh.. but can you do this? I don't expect the rest of you to run with no buffers, as I choose to do. And, I know, for some of you, THAT would be too complicated. Let me explain... Phil... I could not agree more. BUT, there are some here that "hoard" WUs by setting their buffers too high. If a WU is going to get completed in 10 days, but 2 people that get that WU and are each working off the bottom of a ten day buffer, that WU may never get completed by those 2 "confused" persons and possibly passed off to die in our custody. Solution - The system keeps track of average turn around time on a per-machine basis. (Kyong) Give WUs to like-minded buffering volunteers. I HATE seeing a WU waiting for all the "Chucks" and "Administrators" that have an average turnaround time of 6 days or more. MY AVERAGE TURNAROUND IS LESS THAN 3 HOURS! Therefore, I want to get a WU match with someone WITH SIMILIAR turnaround time, or at least, someone whose turnaround, is less than a day. Give the "confused" person with a 5 to 10 day average turnaround per WU to another 5 to 10 day "confused" person. We just need WUs going to like-minded bufferers. Some volunteers may grumble that they need a larger buffer so they do not have down time when the site is down or when we are waiting on WUs. To those users, I say, "GET A BRAIN!". YOU NEED TO HAVE A BACKUP PROJECT! This is a wonderful site run by a GREAT admin, but there are some limits to man and machine and there will be those time that you should be ready to run another project as a backup plan. Since I have a dozen machines crunching here, I have a greater investment than most in what "down" time would cost. Yet, I have no problem using priorities to have ALL my machines still crunch when WUs are lean here for some reason or another. AND even though I do have so many machines crunching, there is NO REASON for me to get nearly a thousand WUs waiting for validation. Makes Phil and I want to reach through the screen and "strangle" you "confused" persons here that cause "cancelled by server" messages when we have done/are doing the work. Also, if I were to buffer my machines for 10 days - averaging one WU every 3 minutes - , and the system just came up, it would not help when I suddenly send almost 5000 completed WUs to Kyong's poor Asteroid's system... with EVERYONE else doing the same! Right now, as set up, I would only be sending 60. And Again... Just coming off a team event for crunching a 15 day PrimeGrid challenge, and although back at Asteroids now for quite a few days, almost all my WUs go to "waiting validation" because of all the "confused" buffering, and will not show a true reflection of work accomplished here for almost 2 weeks. A proper working BOINC setup needs little if no buffering, But if you feel the need, for some reason, set your days for MAX buffer to just 3, or 4, at the most. Any more than that and YOU ARE JUST CONFUSED! AGAIN.. I don't expect the rest of you to run with no buffers, as I choose to do. And, I know, for some of you, THAT would be too complicated. James Lee, A loyal Kyong and Asteroids fan and cruncher. |
Send message Joined: 8 Sep 13 Posts: 71 Credit: 15,807,240 RAC: 0 |
My single computer now only runs work for Asteroids at home, if the project goes down through lack of work units, or some other problem, then my computer is idle until such time that the project is up and running again, I run my computer 24/7, and I only have a buffer set for 0.75 days worth of work units. On average my computer has no more than 44 W/Us ready for crunching at any given time. I agree with the posters about some that hoard W/Us in case the project goes down, but a total waste of time if such numbers of W/Us get timed out, if people are hoarding that amount of W/U`s, then its no surprise to me that the project keeps running out of W/Us so quickly. Hobbies...Visual astronomy and astro imaging |
Send message Joined: 11 Jan 14 Posts: 5 Credit: 87,922,570 RAC: 0 |
The real problem are the "challenges" where teams compete to amass the greatest "score" over the fixed period of such a challenge.... "Bunkering" is a technique to download a vast store of WU's to process within a deadline. Obviously, a large number of these WU's may not be processed in time. Is it wrong ? No the work has value to the project, but the wasted server resources may be. It is not fair to expect all users to have cutting edge machines running 24/7. Obsessive enthusiasts may devoted large amounts of time and money, but we don't have a right to expect all users to do so. The project should encompass all work done as valuable. If a child in a developing country is inspired by an ancient machine processing a single WU then it may change it's life. All IMHO, dunx |
Send message Joined: 8 Sep 13 Posts: 71 Credit: 15,807,240 RAC: 0 |
Well, i`ve had a change of heart, now running 2 computers 24/7, with a 3 day cache, given the chances of running out of W/U`s is often a given problem, i want to be sure my computers are not running idle until new work units are ready. Hobbies...Visual astronomy and astro imaging |
Send message Joined: 28 Sep 13 Posts: 29 Credit: 120,331,090 RAC: 14,573 |
Last modified: 15 Jun 2017, 1:00:43 UTC Tom, What you are doing is perfectly reasonable. The ones that we REALLY hate are the hoarders that HAVE AN AVERAGE OF OVER 10 DAYS per task/WU! Nice of them to contribute - but at the expense of someone that has just completed a task, and then get NO CREDIT as it is "cancelled by server". Since I run 12 machines 24/7 with a turn around time of only 4 hours, I get these "computational wizards" hand-me-downs that have been re-assigned, but theirs just squeak in in time to have my task cancelled after completion. There is NO reason to have a turnaround time greater than 9 days. This causes duplicate work with NO CREDIT! They should be banished! As this is a standard of what happens when a task is re-assigned, NOT the exception - so almost every task that "they" complete, another is cancelled by the server. |
Send message Joined: 11 May 13 Posts: 3 Credit: 6,481,920 RAC: 0 |
Last modified: 9 Jul 2017, 17:45:43 UTC I don't see the benefit in hoarding units unless A. All you crunch is Asteroids@home and have zero power saving features (i.e. fixed voltage, overclocked, NO C-STATES) B. it's the Pentathlon and you want to build a massive bunker C. your internet goes down days at a time regularly |
Message boards :
Number crunching :
Ghost Crunchers / computers