| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: ARP1 workunit *serious* UPLOAD problem since last night! Deadlines may need extending. |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 34
|
|
| Author |
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1326 Status: Offline Project Badges: 
|
Or in other words, they are totally happy with mediocrity. Why strive for excellence when you can have good enough. Or perhaps they would like to provide an all-round better environment but due to circumstances outside of Krembil's control Krembil is providing the best they can. I think they are doing a wonderful job. Keep up the great work  |
||
|
|
wujj123456
Cruncher Joined: Jun 9, 2010 Post Count: 38 Status: Offline Project Badges:
|
IMO the deadline needs extending even without upload issues. Lately I've been getting tasks with less than 2 days in deadline and it takes me 24+ hours to finish crunching on one of my weaker hosts. If such tasks go to people that don't run their hosts 24/7, it's very possible it would time out before finishing. Now let's add the time wasted in download and upload, we'd be timing out a lot of tasks just to resend them elsewhere, wasting even more bandwidth when the server is already struggling.
The project needs to optimize for reducing wasted bandwidth given the clear resource constraint. Hard limit number of tasks per host and in total so that all issued tasks can finish downloading/uploading instead of wasting bandwidth on partially downloaded/uploaded ones. Also extend deadline to avoid eager resend. That's all assuming the team is actively watching the project though, which may not be the case given the silence. The problem here isn't exactly subtle. |
||
|
|
rich@rjassociates.ca
Cruncher Joined: Nov 29, 2022 Post Count: 18 Status: Offline |
wujj123456 said:
>IMO the deadline needs extending even without upload issues. Lately I've been getting tasks with less than 2 days in deadline and it takes me 24+ hours to finish crunching on one of my weaker hosts. If such tasks go to people that don't run their hosts 24/7, it's very possible it would time out before finishing. Now let's add the time wasted in download and upload, we'd be timing out a lot of tasks just to resend them elsewhere, wasting even more bandwidth when the server is already struggling. ========================= Yes, I now have 11 in the upload queue and 7 processing on an i9 149000k, so its not a slow machine and runs 24/7. I have seen tasks with <2 day deadline, not practical at this time. Richard |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 2173 Status: Offline Project Badges: 
|
IMO the deadline needs extending even without upload issues. Lately I've been getting tasks with less than 2 days in deadline and it takes me 24+ hours to finish crunching on one of my weaker hosts. If such tasks go to people that don't run their hosts 24/7, it's very possible it would time out before finishing. Now let's add the time wasted in download and upload, we'd be timing out a lot of tasks just to resend them elsewhere, wasting even more bandwidth when the server is already struggling.. People should not run ARP1 on hosts that do not run 24/7. The project constraints are clearly mentioned.Yes, those 36h deadline tasks are kind of nonsense, considering that I had at least half a dozen hosts just tried to upload finished ARP1 WUs since Friday, that's for more than 3 day alone. And Savas (the WCG team member that sporadically posts now) mentioned that deadlines will be extended, and I can firm in a few cases that they indeed have. Ralf |
||
|
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 823 Status: Offline Project Badges: 
|
I have two machines that run 24/7, left for a long weekend, found them both bottlenecked and stuff due today and yesterday are still struggling to upload. I can manually click the "Retry now" button for this machine but will be another 12 hours before I can do so on the other. Please do extend deadlines or limit resends of work for the time being.
---------------------------------------- |
||
|
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 823 Status: Offline Project Badges:
|
I've been watching as the WUs slowly upload and noticed resends have not been sent out (good!). My stuff is coming in, some of it 36 hours late, and getting verified soon after. Much appreciated! Thank you!
---------------------------------------- |
||
|
|
txt.file
Cruncher Joined: Oct 5, 2023 Post Count: 5 Status: Offline Project Badges: 
|
Again I check my crunching computer and find WorkUnits past deadline stuck in upload.
If I want to play an clicking game I play CookieClicker (or similar). No need for WCG to act as a clicking game. Without manual retry upload BOINC only tries every 2 to 5 hours to upload failed transfers. If BOINC tried every minute the problem wouldn't be that big. |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 2173 Status: Offline Project Badges:
|
Without manual retry upload BOINC only tries every 2 to 5 hours to upload failed transfers. If BOINC tried every minute the problem wouldn't be that big. Actually, that problem would even be worse, as there would be more connection attempts in a shorter amount of time. Actually, uploads (to some degree) and downloads got better during the day... Ralf |
||
|
|
txt.file
Cruncher Joined: Oct 5, 2023 Post Count: 5 Status: Offline Project Badges:
|
that problem would even be worse, I am not sure what is worse: a) many connection attempts cause retry every minute instead every 3 hours b) sending same WU to new computer cause upload was not finished before deadline In my understanding of computer networks connection attempts are pretty cheap to block while doubling data transfer is much more intense. |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 2173 Status: Offline Project Badges:
|
that problem would even be worse, I am not sure what is worse: a) many connection attempts cause retry every minute instead every 3 hours b) sending same WU to new computer cause upload was not finished before deadline In my understanding of computer networks connection attempts are pretty cheap to block while doubling data transfer is much more intense. Ralf |
||
|
|
|