Project will be closed at 1st of February. I see fatal limitations in current realisation, and I think there is no sense to continue doing it in that way. I'll report results soon.
I'm going to do new project, with longer workunits (3-5 minutes each), multiple threads support, data validation, and other interesting things.
Main reason of closing is checking speed. Currently we checked 3.3 million numbers and found 242224 primes in 5 days, starting from 1. Too small. I can compute that amount in minutes in my PC.
Current bottleneck is too short tasks. Server is overloaded with queries, clients are underloaded. And it cannot be fixed in current realisation.