Posts by UBT - Timbo

Mersenneplustwo home page
log in
1) Message boards : Number crunching : Hi - FB Sprint (Message 501)
Posted 542 days ago by Profile UBT - Timbo
3. To enable server processing of completed results now would simply mean that 157,000 results will be past deadline. To extend the deadlines as a one-off would mean that some of them would count but, with the Transitioner Backlog now showing 36hrs and only 32hrs remaining on the challenge, I suspect even that would render this Sprint null and void.


I assume that retrospective credits can be awarded by someone running a script to recognise successfully completed tasks that were uploaded too late - and to then award credits...so some good can come out of this...not that this will benefit any team credits in the Sprint "window" - unless Sebastien extends the Sprint of course.

regards
Tim


Hi all

In the last day or so, I was able to upload and report a large batch of tasks that had been downloaded prior to the Sprint starting...but which were not able to be uploaded once the Sprint started, due to the project website not working as expected.

And these do seem to be getting validated and being awarded credits....even 5 days after their deadlines !

Hopefully others are in the same boat and will see some reward for their crunching efforts.

If that is the case then "kudos" to bearnol for perhaps twaeking the server to allow this.

And it's nice to see that the server is now back up and running OK.

regards
Tim
2) Message boards : Number crunching : Hi - FB Sprint (Message 491)
Posted 547 days ago by Profile UBT - Timbo

You write about a DDos ??? There is no any such report !


Hi marsinph

I think you mis-understand.

The comment is not "about" a DDoS problem.

It is about maybe a few people who could make a script, (via the command-line on their hosts) that causes BOINC Manager to "update" this project every few seconds.

This action is "like" a DDoS, as it causes the project server to have to react to the "update" command from each host.

If one person issues an "update" command, any server can deal with that. But if 100, 1000 or 100,000 hosts all do this continuously, the server has no time to do it's "real" job and it has to "react" to 100, 1000, or 100,000 hosts, all expecting a response.

This action can bring a server down...in a similar way to what happened when the Sprint started on Thursday night and maybe 150+ active members all started "updating" their hosts and trying to upload thousands of tasks.

regards
Tim
3) Message boards : Number crunching : Hi - FB Sprint (Message 477)
Posted 547 days ago by Profile UBT - Timbo
1. The wanless project has a history of losing work even running open outside of challenges. Why? Bearnol's 'trusty server' isn't trusty at all & has been known to have downtime in excess of his own 48hr deadline.


Hi davidBAM

I'm not aware of these issues, personally, but clearly allowing this project to be involved with FB had implcations that clearly were not considered properly...

2. Bearnol got no advance notice of being selected for the Sprint. Very shoddy from Seb. Had he been asked, I would hope he would have declined as he would have known that challenges attract competitive crunchers and that means bunkering. Larger projects know better how to govern that to suit their capabilities.


At 12:47 UTC on Thursday 4th June, (just over 8 hours BEFORE the Sprint started and AFTER a large number of tasks were downloaded by members "bunkering"), "bearnol" was advised about the Sprint and within an hour had responded:

"Don’t know how well the system’s going to cope... but I’m doing my best"

So, perhaps bearnol should have kept an eye on the downloads being sent out in advance of the "official start time"...

If needed, he could have made a settings change, (to maybe limit downloads, extend deadlines, etc) but that didn't happen...


3. To enable server processing of completed results now would simply mean that 157,000 results will be past deadline. To extend the deadlines as a one-off would mean that some of them would count but, with the Transitioner Backlog now showing 36hrs and only 32hrs remaining on the challenge, I suspect even that would render this Sprint null and void.


I assume that retrospective credits can be awarded by someone running a script to recognise successfully completed tasks that were uploaded too late - and to then award credits...so some good can come out of this...not that this will benefit any team credits in the Sprint "window" - unless Sebastien extends the Sprint of course.

I guess it's down to Sebastien as to whether to declare this a "cancelled Sprint"...but it is a great shame that what should be a nice competitive weekend of crunching has turned into a waste of time and energy for those involved.


Back of fag-packet calculation: 157,000 taska at about 1.5 hours each = 235,500 hours of crunching. BOINCstats says there are about 150 active members, so that's about 1,570 hours "wasted" computing power for each and every active member. One can only guess at the amount of wasted electricity that has been consumed... :-(...and ALL the data that has been processed, collected and uploaded that have all been "binned".

regards
Tim
4) Message boards : Number crunching : Hi - FB Sprint (Message 474)
Posted 547 days ago by Profile UBT - Timbo
challenge bunkering causes problems on a small project - again

lots of BAWWWWW about "unfairness" to bunkerers and how it makes the project look BADDDDD


This issue HAS happened before and those projects in the past, were voted OFF the FB Marathon and FB Sprint project lists.

So, the expectation was that remaining projects DID have the capacity to deal with a very large increase in uploads/downloads.

And the project admin here should have perhaps realised that...and put in a little effort "just in case". But that didn't happen, so the admin should accept responsibility for that failure.

i'm sorry, but expecting the admin to jump through hoops to help is unreasonable when presented with this behaviour


I would disagree with you - adjusting the server settings as a "one off" to get through this "problem" is NOT "jumping through hoops"...it is a simple code change, which, if put in place 30 minutes after the server started dropping out, (and one assumes the admin would receive a "notification" that something was wrong) so this COULD have prevented this entire situation occurring.

But it has been left, with nothing done and time moves on, tasks go past their deadlines and many users are upset/annoyed that the admin seems to have done NOTHING constructive.

If someone is in charge of a situation, that then goes out of control, then the obvious thing to do, is to rectify the problem as soon as one can.

Shutting down the server is one way, but it isn't the best solution...as it only helps the admin...it does nothing to help the affected users.

And that's where the problem is now...as the project will probably suffer as no-one will want to process tasks, voluntarily and at the users expense, if they are seen to be treated badly.

regards
Tim
5) Message boards : Number crunching : Hi - FB Sprint (Message 471)
Posted 547 days ago by Profile UBT - Timbo
Hi everyone,

Just as others have said, I am disappointed that we are in the current situation and it is obvious to me that some sort of compromise solution should be made to help resolve the situation.

Obviously, people who've spent time and money on hardware and electricity to support this project would like to see a "return" so I cannot see the harm in extending the deadline (though in the current FB Sprint, it might mean that UK BOINC Team loses it's 1st place).

Secondly, it does seem unfair on everyone who has supported this project, that the admin of this project is purposely denying anyone from earning any credits, just so they can let things "calm down" - because the longer this goes on, the MORE damage is being done to the reputation of this project.

And the end result will probably be that every member of every FB Sprint team will simply refuse to process any further tasks from Wanless project.

Thirdly, it does seem strange that the project hardware has not been upgraded, especially since the admin of the project has allowed the project to be included in FB Marathons and Sprints.

So, "bearnol" - please do the right thing for those who've supported you - extend the deadline - and at least keep some people happy.

You should then contact Sebastien and remove the Wanless project from inclusion in their Challenges...that will allow you to keep the existing hardware.

Alternatively, put a link on your home page, asking for "donations" and I'm sure some members might be willing to offer you a nice condition replacement server, or even some $$ to help you purchase a new one. (You seem to have around 154 "active" users running tasks, so if everyone gave you $2-5, you could get a new mobo and memory and even a nice fast SSD - or just upgrade your broadband connection with Virginmedia ?)

But, admin, please DO SOMETHING, instead of referring people to old posts and closing down your server...as both these actions do not actually help ANYONE.

Simply burying your head in the sand is only putting off the issue and it will still require effort on another day - so why not put in the effort NOW?.

best regards
Tim
Founder, UK BOINC Team




Return to WEP-M+2 Project main page


Copyright © 2021 M+2 Group