Re: scaling performance vs number of CPUs

[Scott Koranda]

Thanks for the input.

Yes, the queries are examining similar ranges of data. The data is essentially time ordered on one colum with the other columns being different. Each of the 25 queries is examining the same time range on the time ordered column but sampling differnt values for the other columns. Since the data goes into the table in time order it is almost assured that the queries are for the most part examining similar ranges of data.

I looked into next-key locks and read about transaction isolation levels. I did

SET GLOBAL TRANSACTION ISOLATION LEVEL READ UNCOMMITTED;

hoping that SELECT....FROM... will not set any next-key locks. Unfortunately it did not help, so I think next-key locks may not be the problem. Is there a way I can tell for sure?

I am issuing the requests by simply breaking the 25 queries up into 4 files and then doing

time mysql --user=dbuser --password=PASSWORD < query-set-0i.sql > out.i

where 'i' is 1 through 4 (each query set in a different bash shell).

Any other thoughts?