PDA

View Full Version : Small tests being handed out?


Paperboy
10-10-2004, 11:02 PM
I just checked out my user stats and I see a significant drop in my production for a few hours today and 3 test completed when I only run 2 computers. Are smaller tests being done, and can anyone tell me what k/n I tested earlier today while I was at work?
http://www.seventeenorbust.com/stats/users/user.mhtml?userID=297
Joh14vers6
10-11-2004, 01:26 AM
Why come dropped tests from the garbage,secret, supersecret, 31337 queue come all in the droppedtestqueue, so that the normal users get this work instead of garbage,secret, supersecret, 31337?
vjs
10-11-2004, 08:38 AM
A dropped test is a dropped test it doesn't matter where it comes from or the n-value. Your chances of actually getting one of these low n tests are actually pretty rare, however it is impossible to get a very large n-test (n=~13.6m) using a standard account.

If you search the message forum you will see how and why this happens. Personally if I could do low n-testing without too much trouble under my username I would I think it's just a matter of taste.
Joh14vers6
10-11-2004, 11:33 AM
Originally posted by vjs
A dropped test is a dropped test it doesn't matter where it comes from or the n-value.

I know. What I am asking is actually, is this the rightway how it should work?
Mystwalker
10-11-2004, 04:17 PM
One could argue whether it should be inserted back into the "second-pass" queue - but I think there are more important problems with the queues... :(
vjs
10-11-2004, 04:54 PM
Joh14vers6,

I think Mystwalker summed it up best...

I wasn't sure if your question was

"should dropped tests from , error fix, garbage, etc. ques be put into the "normal users/dropped tests" que when dropped?"

Probably not, but it has been setup this way and is working "correctly".

Should they be routed back to the orginal que they came from?

The answer is probably yes here.

But the amount of effort to rewrite the server queing code is greater than any lost CPU effort. Anything that was dropped from these other ques would have to be done eventually anyways, so having them done by the normal effort is better than not having them done or having to rewrite the server que...

Also I think the chances of finding a prime on a double check n is really not that bad comparred to the current n, when you consider you can test ~100 double checks in the time it takes to do one n=7m.