by We have a need for a very "fast" hardware locking mechanisms for a high
speed OLTP system. This mechanism must not only be very "fast" (several
thousand locks/unlocks per second), but it should be scale well. A
typical multi-cpu machine will have about four to six processes
(multi-threaded), all accessing/setting these locks. We're getting
conflicting stories from about which is faster, System V or POSIX
semaphores under Solaris. If anyone has crossed this path before, I'd
appreciate any information, if benchmarks have already been done, I'd
appreciate it I can get a copy of the code used during those benchmarks.
Thanks in advance,
Roque Solis
Systems Engineer - The SABRE Group