> Any suggestions? If you know that you will likely only perform per-site queries then you want all the readings for a given site contiguous in the database file (or files). You can accomplish that in many ways, as you've outlined.
Hopefully your reading_id's always increase as time goes forward. Consider collapsing timestamp and reading_id into one value (timestamp) and make that your primary integer key if you can. No point guessing about the various strategies - try all your ideas and do timings using typical queries under normal usage patterns (vacuumed vs. no-vacuum). There are benefits to seperating the data into seperate databases and/or tables (reducing row size by eliminating site_id) as well as keeping it all together (reducing your admin/coding effort). In general you want to reduce the reading row size to a minimum number of bytes to obtain greater insert and query speed. The manner and order in which you populate your tables also plays a large role. If you populate data from all sites each day into a one-database/one-table solution without vacuuming then data from each site will be farther apart in the database file, and your per-site queries will take longer. If you bulk load all the data at once prior to analysis do so one site at a time with rows inserted in order of time. That will reduce or eliminate the need to vacuum the database. ____________________________________________________________________________________ Need Mail bonding? Go to the Yahoo! Mail Q&A for great tips from Yahoo! Answers users. http://answers.yahoo.com/dir/?link=list&sid=396546091 ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------