Re: checkpoint time > 15 sec.

In article <36963598.4B7D@bloomberg.net>, Art S. Kagel <kagel@bloomberg.net> writes >may become a larger issue there also. Because of the algorithm that >selects LRUs it is certainly possible for all four CPU VPs to hash to >the same LRU on their first attempt causing 3 bufwaits before the >second VP gets the latch from the first and the other two timeout their >spin cycles and rehash. > ?? Surely the hashing algorithm needs fixing then? Even a simple one like hashing on which CPU VP it is would yield better results.. i.e. Each CPU VP gets it's own set of LRU's. >Art S. Kagel -- David Williams