A lot of things have changed since I made the first CLQ 10 years ago. Most noticeably all kinds of anti virus and anti hack stuff has been added to Linux that gets in the way if you want to introduce massive parallelism.
I could work around this with a customized Linux kernel, but choose to work around this instead.
I now have a working "superscalar" scanner that can scan 1304 active UT2004 servers in 5 seconds. CPU load is about 5% for this, it scales nicely so all cores are used. All scan data is preprocessed, bzip compressed and stored on disk. This will allow me to achieve the 15.000 servers I wanted to scan per minute with great ease. I might even increase my ambitions and scan servers more often than I intended.
There is still a lot of room for improvement. It is highly configurable allowing me to change it in the future if I start scanning larger amounts of servers. Also the scanner is not connected to the processor yet (which might be caused by the fact I did not build the processor yet

) so the database is not optimal yet.
The amount of data needed for this is 4 megabytes (compressed data). This means a single server scan takes up 3 kilobytes. This means the 1 TB disk can store 330.000.000 server scans. With 15000 scans per minute this means I can store 22000 minutes (or 15 days) of data on that disk. Hm, that is not very much actually. OTOH I do not really need all that history, I only need to store the resulting statistics. Oh well, I will figure out this part later on. I negotiated 2.25 TB of traffic per month for this with my provider, didn't realize I might actually use it all

.