| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home
Thread: Dual Boot Machine Missing Deadlines Due to Long Work Units, Double Load |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have been running grid computing software on my computers for at least a decade but only joined World Community Grid five or six weeks ago.
My current computer is a dual boot Pentium-4 Dual Core 2.8-ghz with a gig of memory, and 70-gigs of free hard drive space in XP and 27-gigs of free space in Ubuntu Linux. I usually use the computer between 7 and 8 hours a day, and choose the OS based on what I plan to do. For writing, email and web surfing I prefer Linux, but there are Windows programs I need to run in Windows and I spend most of my time there. Thankfully, the BOINC client seems to run fine in both OS's. The problem I have encountered recently is with the long work units from FightAIDS@Home. They run fine, but take a long time to run and I have been missing deadlines and getting messages that I might want to "abort" the work unit. It doesn't make sense to me to dump something the computer is 75% finished with. The question is: Do these work units become useless if the deadlines are missed? Another problem is because I am running dual boot, I can't be running one OS when I am running the other since it is the same machine. I have been assigned long FightAIDS@Home work units in both OS's, and with deadlines approaching I will try to stay in the one closest to deadline, but when the task is finally done I switch over to the other OS and find that deadline is too close to meet. I currently have two FightAIDS work units within an hour of finishing in Linux and have opted out of being assigned any more. I have at least one FightAIDS work unit in Windows that is due in a few days. All the projects on the community grid seem worth the effort, and I don' care what my machine is working on as long as it is making a meaningful contribution. If it is constantly missing deadlines it can't be making a maximum contribution. I don't care about points, only productivity. My question is: Does it make sense to opt out of the long FightAIDS units and concentrate on shorter unit projects? I assume faster quad core computers running one OS might be better suited to working on the AIDS project and they might be assigned more if I cover the other projects. I know we are talking minor percentages but in the end every cycle counts. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yes, if you are having deadline troubles, then you should select the projects with shorter work units.
By the time you miss a deadline, the work has already been reallocated to someone else. You may still get credit, but there is a good chance the work is wasted. Nutritious Rice for the World has work units that run for 8 hours regardless of your computer speed, and the deadlines are generous. Sekerob has a graph of current work unit sizes linked in his signature. Sekerob? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello CaptainMidnight,
It sounds as though you do not want a large cache of work units. I suggest that you select My Grid - Device Manager - Device Profiles then select your BOINC profile and set the Workunit Cache Settings to: Connect to network about every 0.1 days Cache 0.0 extra days of work This will keep you down to only 1 work unit until 2.4 hours before you finish it, when a new work unit will download. Lawrence |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
Hello CaptainMidnight,
----------------------------------------My "colleagues" have given you two essential advices for solving your deadline issues. I will give you the third one, valid for anybody who needs to stop Boinc sessions once or several times a day: choose the projects with the shortest checkpointing periods to lose minimum runtime when tasks are restarted. Currently the projects with the shortest checkpointing periods are the Rice (NRW) and the Dengue (DDDT) projects. On a machine like yours they should take a new checkpoint within less than two minutes after the previous one. The Cancer project (HCC) will checkpoint every 6 to 10 mn on your PC, and it has short WUs (about 4 hours on your PC). Make sure that your Write to Disk interval is set at the 60 seconds default to get full benefit of a frequent checkpointing capability. I am swapping a quad running at about the same speed as your PC between XP 32-bit and Ubuntu 64-bit. I have selected DDDT under XP and HCC under Ubuntu, with 0.1 day of total work buffering and I have no deadline problem, even when I receive "repair" WUs with a shorter deadline. Cheers. Jean. ---------------------------------------- [Edit 1 times, last edit by JmBoullier at Jun 29, 2008 11:15:46 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Gentlemen: Thanks for your quick responses and good advice. I made all the preference changes you suggested and have watched closely for the last 48-hours to see how things worked out.
In both XP and Linux everything is running smoothly and I have already been able to return a couple of shorter work units. Since I only opted out of the Aids work units, I was assigned a long 20-hour unit for Protein Folding but because the deadline is three weeks away I don't expect that to be a problem, but I will keep a close watch. Thanks again for your help. Bill Scheffler aka, on this board only as: Captain Midnight  |
|||
|
||||
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
Hi Bill!
----------------------------------------I have hesitated on writing "stay away from HPF2" in my post because of its "monster" WUs. But since the return delay is quite long I did not.  However if you are running rather short sessions there is another reason why you might want to avoid HPF2: the checkpointing period for this project is much longer than for others, and it goes increasing along the computation. In the last third of a WU you might lose up to one hour or more if you close a session just before the next checkpoint is taken. Cheers. Jean. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Is HPF2 or HCC displaying this ever extending checkpoint time profile? [a brain-jam as I thought it was latter and as such wrote it up that way in the Start Here FAQ some time ago]
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
For HCC, yes, the checkpointing period is varying and it is usually longer at the end of a WU than at the start. But it does not matter much since it is usually varying between 6 and 10 minutes (for a total runtime of 3 hours), which is still very manageable and avoids losing much work if a session is stopped.
----------------------------------------For HPF2 I have not run one under Boinc for a long time, but the last time I did it it was perfectly obvious, and on the P4 at 2.4 GHz I remember having seen up to 75 or 80 minutes between checkpoints at the end of some WUs. Which, for me, is far too much. Cheers. Jean. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Yes, if you are having deadline troubles, then you should select the projects with shorter work units. By the time you miss a deadline, the work has already been reallocated to someone else. You may still get credit, but there is a good chance the work is wasted. Nutritious Rice for the World has work units that run for 8 hours regardless of your computer speed, and the deadlines are generous. Sekerob has a graph of current work unit sizes linked in his signature. Sekerob? The hits on that Supplemental Graph page in the Start Here forum is not even 1 per day so my absence did not inconvenience that many.... and in case there is still interest... the FA@H monster work units are behind us again, for now:
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|