from the archives October 2005
I named the big one Potter and the little one Hagid. We're running benchmarks from Snape. Potter is a temperamental one, doesn't like my 32bit text editor and won't synch IDs from my database to the Deployment Server, so that when the applets go a-calling, they get rejected out of hand. But Potter hasn't given me a real fit like Hagid did this evening. Three tests. One gives 40% errors I can't explain. The next one goes flawlessly. Then I change one little parameter and the whole craptard goes zombie. This, ladies and gentlemen is why NT sucks, no matter how hefty the hardware. I've got 300 databases running simultaneously and the agent freezes. Can I kill the zombies all at once? Noooooo. NT has no such animal as a kill -9 (the lung ripper). It just sits there and divvies up the full weight of 8 3GHz Xeons burning 100% across 300 processes that I've got to kill one by frikken one. With a mouse and four clicks. Warning, killing this process may cause system instability. Are you sure?
Anyway. Next time I'll keep Load Runner throttled and not try to launch 300 databases all in the same second. Lesson learned.