Could it be possible to take a few thousand work units off of a server for reservations somehow (eg. PPSElow, etc. where the workunits are in the millions), so people can tear through them faster? I can run PPSElow WUs on PRPnet at 1 per 90-120 seconds, yet using PFGW and fully primality testing it takes 55 seconds (on the same computer). If we had the chance to tear through WUs faster, then we could have a chance to find more primes. More WUs = higher n's. Higher n's = larger primes.

Could this be possible? The candidates would just be in a text file, say 1k to 100k candidates at a time for testing with PFGW - and I don't see an issue with just having a program writing thousands of candidates to a text file.

I think you could do this by just allowing more than 25 wu to be requested at a time. In the last run of PPSELOW the server was overwhelmed by all the requests and I saw frequent timeouts (sometimes to the point where my wu expired...).

A large part of that was bkus had almost all servers setup to pull 5 wu, if he had done 25 at a time it would have been a lot less load on the server. That being said he's moved onto Boinc after I lost our race to the top of PPESLOW wager :P (On a positive note I've got him hooked on primes)

I know that I can request more than 25, it's just that I want to be able to reserve a few thousand work units to be useful OUTSIDE of PRPnet. PFGW is a lot faster, and thus should be used for PRP/primality testing.

If I want to reserve a few thousand workunits, I have to set up about 32 clients on my 4 core system, then manually go in to the work files and copy out the k/n pairs, put them in to a text file and PFGW test it. If we had generated text files with the pairs in and a reservation page alongside PRPnet, it would make for much faster testing, wouldn't it?

Oh, you can use pfgw already, just uncomment your llr= line in your .ini and it will use whatever pfgw you have specified for PPSELOW. I didn't find a big difference in speed between llravx and the latest avx build of pfgw.