I just wanted to apologize to everyone for the confusion regarding the range in the server for this drive.

This preliminary very undersieved n=2K range was an extremely unusual situation that only came about as a result of the rapidly rising 5000th place prime.

It is my hope that not too much inconvience nor lost CPU time was had by all involved.

If there is one thing that we found out by accident here, it is certainly how robust David's servers are, and in more than one way:

1. It can handle 100's of cached results dumped all at once on it (as long as it is not from a Proxy server).

2. It can handle the fast testing times of a lower n-range.

Had we not tried the n=200K-210K range for a while, we would not have known #2 so some good has come out of it.

This makes me wonder if his servers couldn't handle candidates as low as n=50K with testing times as low as a few secs. We'll be setting up the 9th drive for n=50K-350K in a couple of days. If David or others think that may be worth a try, perhaps we can set up some sort of testing for that; perhaps with one low-weight k to start with. We'd probably set up a completely different server for it. I'm thinking we'll use port 8000 for n=352K-500K after sieving is complete in < 3 weeks, although that can be discussed.


Gary

[quote=gd_barnes;156648]...

This makes me wonder if his servers couldn't handle candidates as low as n=50K with testing times as low as a few secs.
...
Gary[/quote]
Maybe have a rule e.g. MaxCoresPerPerson = n/10,000?

For me, I'd rather run on servers than do manual ranges.

[quote=Flatlander;156650]Maybe have a rule e.g. MaxCoresPerPerson = n/10,000?

For me, I'd rather run on servers than do manual ranges.[/quote]


Very good idea on the rule. We probably need to add some sort of caching rule too. Although his servers could handle 100's of pairs dumped at once at n=350K, doing so while at n=50K while handing out pairs to everyone every few secs. (vs. 110-150 secs.) might be a different story.

For the most part, I also prefer servers but am a bit annoyed by the 5-10% testing time loss of LLRnet. Alas, that will hopefully go away with the new PRPnet software that is currently being tested at CRUS. Very exciting stuff!


Gary

LLRnet IB8000 has completed 350K-352K, lresults emailed to Gary. :smile:

Reserving n=352K-360K for port 8000.

As soon as we get the word from David that this file has been loaded, the 8th drive will have officially restarted.

This should be an exciting drive with many top-5000 primes coming more quickly than the 5th/6th/7th drives. Don't worry about the already-known primes. A large percentage of this range is unsearched.

Some manual files have been posted in n=500 blocks.

I'll post known primes after the top-5000 site comes back up.

Let's rock! :smile:


Gary

Port 8000 loaded, server running, using auto-notify if primes are found.

Reserving 360.0-360.5

Reserving n=360.5K-375K for port 8000.

With the current range completing in < 1 day :surprised and the rally coming this weekend, we'll load a big file that should last through the rally.


Gary

Loaded up.
Where are the Rally notifications?
I can't find it here or over on Free-DC...
When does it start? How long will it last?

[quote=IronBits;159533]Loaded up.
Where are the Rally notifications?
I can't find it here or over on Free-DC...
When does it start? How long will it last?[/quote]

Huh? I just posted something in the rally thread late yesterday asking if people would post their # of cores on it. See the 1st post of that thread. It's stickied so easy to see.

I also sent a PM to Lennart letting him know about it.

Loading up for the rally:

Reserving n=375K-400K for port 8000.