problems with garbage acct.

could the number of residues recieved for a test be added to the garbage que list. This would make it alot easier to see which tests are for the main prp and which ones are abandoned tests that have been p[icked up again. Would it be possible to have someone who uses the garbage que to only recieve a tests that has not had any residues turned in yet? and have only one of each test handed out in garbage (i don't know how it is now but garbage used ot give everyone who chose to run it the same test. ) Not that i believe people should be running garbage acct. but the way it is now anyone running it is performing a triple or quadruple check on the N values that are obviously not prime. also the original purpose of the garbage que was to make the bottom limit of the min n vallues where a prime could be move faster. With the way its being run now it does not achieve this because all of the low tests being handed out (around 4 million) have been completed by someone else along time ago. (And they already have a double check in the works which is what you're looking at, be it intentional or not.) I don't want ot see the garbage acct removed because i liked the original plan for it. I don't like seeing the n lower bounds sitting in one place for many months because there is one person running a test on a P1, or because someone is intentionally holding on to a test for an obscenly long time by only submitting blocks once a month just to have this bottom value. (I'm pretty sure it has happened) If someone had been doing that and the values turned out to be prime the prime could be found months if not years behind scheduale leading to maybe months of wasted effort.

Quote:

---

Not that i believe people should be running garbage acct. but the way it is now anyone running it is performing a triple or quadruple check on the N values that are obviously not prime

---

I am confused,

in this post http://www.free-dc.org/forum/showthr...&threadid=7929

it says that garbage is very useful??

Keroberts1,

Are you sure this is the way garbage is working right now, if so it doesn't make sence and I agree. Let me point out some of the aspects of the way I thought garbage, and supersecret work.

Supersecret obviously is intended for double checks and pulls k/n from the second-pass que, once a k/n pair has a matching reside it's considered a complete double check. If supersecret returns a non matching residue it should be added to error fix que.

Garbage was initally intended to pull k/n tests from the garbage que, this que contained k/n pairs that may be on going tests... in any respect no residue, complete test has been done.

Secret at one point was the account that pulls from the error fix que.

Now, if garbage does a test and completes a residue this test is eliminated from the garbage que, correct? Or at least it should... I suppose that the logic may not be correct and it sends the k/n pairs several times since it is the lowest non-residued k/n pair even if assigned to garbage, I think this is what your getting at, correct?
If so this is an issue.

I think the major problem with these que system currently is garbage pulls from both error fix and garbage simultaneously...

If I could redo the que structure I'd do the following.

First create one more account errorfix
and one more second pass que for n=3m+ like before

Have,
supersecret pull from second-pass
secret pull from 3m+
garbage pull from first, garbage then dropped tests, and make sure removes the test from that que once assigned, if it's dropped it goes into dropped tests.

now the new errorfix account would first pull from error fix, then garbage by lowest.

Also I don't think P1'ing a test keeps the k/n pair alive on the server, I don't see how this happens since no blocks are returned.

Also if someone is holding on to a test for a very long time this is fine, just test it first with garbage before they finish. If they finish before second-pass great if not we dc it before and they submit a tripple check.


Also I really want to point out that a 3m test que is probably a good idea at this time, people with fast computers could run it insead of 1.3m tests... Currently fast computers have either the choice of main que which they may not want to do or garbage 4.5m. It would be nice to have a happy medium.

AFAIK, tests that were "overdue" were put into the garbage queue. So far so good. To be honest, I haven't read anything about tests handed out several times by the garbage queue yet.

The question for me: Is this queue updated over time?

Which means:

- When the original tester finally submits a result, will this test be taken out of the garbage queue (assuming it haven't be handed out before)?

Note: This is not that crucial, as not doing so would lead to an early double check, which most likely has to be done anyway.
It seems like this isn't done, as often the next test in the garbage queue (currently: n = 4533970) is not listed in the Tests That Are Over 90 Days Old - table...

- Will a factors lead to the test taken out of the garbage queue?

- Does the queue get restocked on a regular basis?

all tests below 5million somehting have residues. The garbage que does not heck for residues. It simply assigns tets that are over due whether they have been completed before by someone else or not.

I thought garbage was just for Louie to run - ie it assigns the lowest first-pass with no residue, even if it is still being processed (if slowly) or if it has just been given out to someone else running as garbage - purpose to keep the range ticking upwards. Have I got the wrong name here? Or did the behaviour change?

These are open accounts for everyone

garbage, secret, supersecret,

and 31337 which pulls from "largest-prime"

There are 4 or 5 other accounts which are not public.
These 4 or 5 other accounts do specific testing and are just for Louie et al.

One of those accounts pulls from the "residue-recovery" que.

Ques can be seen here http://www.seventeenorbust.com/secret

With that being said, looks like this thread hasn't gone unheard!!!

This will probably only confuse people more, but here goes.

The "secret" account draws from the "second-pass" and "error-fix" queues. So it's used for people who want to do only double-checks. Right now the double-check threshold is controlled strictly by how many people run clients on this account. There's not (yet) any balancing system in place to keep the second pass threshold at any specific proportion to the first-pass threshold.

The "supersecret" account is in fact identical to secret. It had a different purpose in the past, but not anymore. It's a double-check account just like "secret."

The "garbage" account draws from the "garbage" and "dropped-tests" queues. Its purpose is to narrow the window between "highest n inclusively tested" and "highest n currently being tested." In other words it tries to clear up low-n tests that got dropped and are currently limiting the inclusively-tested threshold. It used to be populated automatically, but honestly I'm not sure if this is still true. I can't seem to find it if it is (there's nothing in crontab or in the server code that touches this queue). I'd have to check with "et. al." to figure out what happened to this.

The "dropped-tests" queue is exactly what it sounds like.

It's true that there are a couple other special accounts, but don't worry about them. Pretend they don't exist. (Used for testing and also for tests that, for whatever reason, we want to test ourselves on our own machines rather than handing them out to the world.)

Signed,
"et. al."

AFAIK, the "garbage" account also pulls out tests from the error-fix queue.

Yes garbage pulls from error-fix and garbage which ever is lower.

Secret pulls from double-check and error-fix which ever is lower.

These should be reset!!!

Garbage only garbage,

Secret error fix, or garbage which ever is lowest

Supersecret, double-check, error-fix, garbage, whichever is lowest.