On Sun, 21 Sep 2003, Brion Vibber wrote:
See benchmark:
http://meta.wikipedia.org/wiki/Lock_response_times
http://meta.wikipedia.org/upload/1/1d/Lock-times.png
Of course, I could be doing something wrong, and this benchmark is very
very naive, making no attempt to work with different types of requests.
As I argued, using different types of request would probably give a worse
score for strict locking, because the faster processes have to wait for the
full execution of the slower ones. With two processes, there is loss rather
than gain even in the theoretical case when the slower process takes more than
3 times as much time as the faster one (if one process has time 1 and the other
time a>1, concurrent execution will give execution times of 2 and 1+a, average
1.5+0.5a, sequential execution will give average execution times of 1+0.5a and
a+0.5, average 0.75+0.75a).
Andre Engels