F
frank.meulblok
Guest
[quote user="George Potemkin"] > Reading _Lock table is a recipe for a performance disaster. Not an issue since 11.3 [/quote] Since 11.4 actually, with further tweaks in 11.5. In 11.5+, assuming a sports2000 database started with -L 500000 a very basic benchmark: ETIME(TRUE). FOR EACH _lock: icount = icount + 1. END. MESSAGE ETIME icount. Will give sub-second response times under most circumstances. If you're using remote connection and actually have hundreds of thousands of active record locks you might see more human-noticable response times - total size of lock table and number of actual locks used are still the main factors in time to execute. Same query in 11.3- probably will have you kill your session before it completes, as you'll get tired of waiting before it finishes.
Continue reading...
Continue reading...