G
George Potemkin
Guest
Thanks, Jeff. The pica queue lock is the same as service latch lock, isn't it? > So the lock is held for a very, very, short time in duration. Is it less than a microsecond as in case with the "spin" latches? Sorry I don't know a correct term to name the classical latches (like MTX, USR etc) in opposite to the administrative latches (like the pica service latch). Access Count: 150,000,000 per 30 min in my example. If access count = lock count then we can roughly estimate the total duration of the queue lock as being less than 150 sec during 30 min interval or 8% (or less) of all time the queue is locked. Is it correct? Access Collisions to Access Count ratio is only 0.017% in my example and 0.019% in Dmitri's example. Which processes and how activily are competing for the queue lock? Why the collisions are so rare?
Continue reading...
Continue reading...