Need help getting highly-optimized inserts for float data

[Travis Hamrick]

Hi,

I am writing a routine that is attempting to save floating-point data from a process that produces measurement data about two hundred times per second. All of the data is either unsigned integers (foreign keys) or floating-point data. I'm saving 2 or 3 floats and one foreign key per row with about 30 rows per table for 4 or so tables each iteration.

My goal is to see if it is possible for this data to be saved directly to the database, without the database getting too far behind.

-- A few selects may be queried while all this is going on, but not on the tables I am inserting to.
-- I know in advance what the foreign keys are, so I'm using the insert into * values (), (), () format
-- I'm using MyISAM for the table engines -- is there a better fit for this?
-- The program is a C++ program, but I'm currently using the C API for mysql. Does the C++ api have any performance advantages for me specific to this situation?
-- Converting all these floats and integers to strings for the query, so the database can just convert them back seems like a waste. Is there a better way?
-- I am looking into using disable/enable keys to disable the keys until the end. Are there other such tricks to use?
-- This burst of data may run for say 5-10 minutes. Any database parameters that can be tuned for this situation

Any other suggestions?

Thanks