OJ Board

Due to some personal problem (Saturday is not a holiday in Bangladesh)
I had to request Prof. Miguel to shift the first regional warmup contest to 5th October (Sunday), 2003 at 9 GMT. Previously it was scheduled on 4th October. I am sorry for this change.

I guess nobody will mind, and I think that I can assume that I speak for the majority of people here if I tell that we're so grateful to you and all people involved in acm, acm.uva.es, and icpc, to organize all this !!

Thanks! And we are also grateful to you all for participating in large numbers. Actually a good problemset and large participation is necessary to make a contest successful.

I also think that you and the whole team make really good job!
But large participation in the last contest (Warmup 1) caused some troubles with the testing system (400 submission waiting in the queue).
Will it be fixed in the next contest?

Just curious: what scheduling algorithm are you using on UVA-OJ?

Last contest there were 2000 submissions. Given 5 hours, that is, 18000 seconds, there should be approximately 9 seconds for each run. But a lot of them finished in less than a second.... how could the system be possibly that slow?

I am also curiuos about that. But, i think, your calculations are not realistic. You have to add the compilation time and the fact, that in the first 2 hours the system was quite idle.

I've noticed that many programs have been running at the same time. I wonder if simple FIFO algorithm could be better for the system.

It is very difficult to make sure that a contest will finish successfully. For example the two waterloo contest and the contest before that ended successfully but the regional warmup did not. It had some problems with high time limit.

To ensure that a contest runs successfully a contestant can do the following:

a) Be sure that his solution is correct. I mean not submitting just after getting the sample right. I will try to put some samples which are critical so that people can understand that their program will fail even before submitting.

b) Wait for the judge response before submitting another one. When the judge response delay is say 5 minutes, if quite a number of contestants submit before they get the response, the queue gets longer and the judge response becomes even slower. So once the response gets slow there is no way to recover.

c) I am thinking about stop accepting submissions until the queue is empty. This may be the only solution to save a slowed down contest.

-Shahriar Manzoor

Yes,that's a good thinking from Manzoor.Onething is clear that there is a large participants during the contest time and sometimes it makes collapse in submission.So this typical problem should be overcome in the upcoming contest.

ALL the best for ACM.

ARNOLD VAJNA