Competition side effects?

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109408994544
RAC: 35176789

The host with the 33 resends

The host with the 33 resends from my previous message has just had a further cache top-up.  This time it was sent 35 tasks none of which was a resend.

Looks like the scheduler has now dealt with the backlog of resends :-).

Cheers,
Gary.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109408994544
RAC: 35176789

Gary Roberts wrote:archae86

Gary Roberts wrote:
archae86 wrote:
I wonder whether the participants will eventually process and report them, abort them, or just abandon them to time out.

Since quite a few of the 'contestants' seem to be rather fanatical about the 'competition', it would make no sense for them to keep any excess around where it could interfere with the ability to download the maximum possible for the next 'event' to follow this one.  I would expect all excess to be immediately aborted.

Judging by the sheer number of resends that keep continuing to arrive, I was quite wrong with that 'expectation'.

I guess I knew (human nature being what it is) that there was probably little hope that people would make some sort of effort to calculate in advance, some approximate limiting number of tasks they could expect to process.  After all, why should they bother not to be 'greedy' when everybody else was bound to be 'greedy' anyway.  And, on top of that, the likely outcome from trying to be considerate towards the system and all the other participants could easily lead to escalating your own chances of 'running out' early if the excessive load caused the system to break.

However, what I didn't expect was that some very inconsiderate people would leave the huge excess of tasks, literally to 'rot on the vine', rather than immediately aborting them.  After all, aborting even thousands of tasks is pretty straightforward and quick to do so why not spare a thought for your many quorum partners and just quickly abort them?  Some of these people obviously have no intention of providing ongoing support to the project and couldn't give a hoot if they are 'trashing the joint' on the way through.

Over the last couple of days, I've started to notice, once again, an increasing number of resends.  On the basis of some people not aborting their excess when the Einstein part ended, I was expecting to see a bit of an increase around 13-14 days after the Einstein phase finished.  This would be for tasks received on the last day of the event but not crunched and not aborted.  But that's not for a couple of days yet so why this early?

I was curious, so I decided to have a look.  I picked one of my hosts and one of its resend tasks received about 6 hours ago.  The quorum for that task led me to this host and its task list.  To summarise, a single resend task received today, picked at random (the only one I bothered to look at) gave me a host with a deadline miss that created the resend in the first place.  On looking through the tasks list for that host, there are very few 'valids' still showing, but somewhat more 'pendings' - both as you would expect - but huge numbers of errors (deadline misses) and huge numbers of 'in progress' that will time out over the next few days.  The last date I can find for a returned result is May 15, 07:47:52, but there would most likely be heaps more later than that that got validated quickly and removed from the online database.

The obvious conclusion is that this host had huge numbers of excess tasks when the event finished.  There are 647 still showing as 'in progress' that will become resends as they expire and 426 as errors because of a deadline miss that has already occurred.  There was no attempt to abort any of the excess.  I doubly don't know what this person was thinking.  Firstly, why get such an excessive amount, much more than you could possibly need, which continued right up to the last day of the event?  Secondly, why not abort this huge excess once the event is over?

We will still be seeing resends from this host, and probably many others, right up to May 29, before this anti-social behaviour is finally over.  I can't imagine that the single task I picked would immediately find the only host doing this.

Cheers,
Gary.

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 437
Credit: 137331180
RAC: 18170

I am sure the projects

I am sure the projects welcome the additional processing power these competitions bring and there probably isn't much they can do about the side effects. Probably the simplest approach would be the project reminds the teams (email the team founders) that they should either complete the work they have downloaded or abort it immediately after an event.

As a longer-term solution maybe the project could have some fancy scheduler setting to set the expiry date when its a member of team that has signed up to the competition. That would take a bit of back-end plumbing work as you'd have to know the start and end dates of the competition and which teams have signed up.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.