A surprising problem that developed as part of the move is that viewing threads on THR is much faster than before, as expected. Posting, however, has become a problem.
The quick solution is probably to get the uber-server I bought in January repaired, move the SQL server to it, and let the additional horsepower it offers cover up the problem. In reality though, the database server is only handling about 70 queries per second during most of the day (maybe 2-3 times that at peak), and this level of service is something we should be able to provide using the existing hardware. Additional horsepower would mask the problem, but not fix it.
I made some tweaks around 4:30 this morning that should help, but as I mentioned in the “Change is hard” post tuning will end up being an iterative process. I’m gathering statistics, and will make some more tweaks tomorrow, and the cycle will restart again.
Things are better. Backups used to take 45 minutes; this morning it took 5, using exactly the same technique. We’re moving in the right direction, but it will take a bit longer to get there.
If you see problems that you think I’ve missed, or that you think might give better insight into what’s going on, please feel free to PM me with the details.
Thanks for your patience.