As I have previously written, we have a use case where a process is continually inserting records into a table at a continuous rate of about 8/second right now, every second of every day. There are test measurement records being received from devices in a network. Because of disk space requirements, only 30 days of data is kept and older data is purged out each day. The purging has a great effect on the rate of insertion and keeping a sustained rate. In addition, since this is a continuous process, there is no down time except maybe once a month.
It seems this use case is not handled that well by the database and search the net, it seems that many other databases have the same issue. So I was wondering if it might be possible to attack this problem from an application level. Instead of having one table that is constantly being inserted and deleted from, what if I had 6 tables, one each for 5 weeks and the insertion always inserts into the proper table for the week associated with the data. When the 6'th week's table starts to be inserted into, the oldest table can be dropped and recreated as the next week's available table. It does make the insertion and query tricky but purging out results is a matter of dropping and recreating a table. Any thoughts on an approach like this? This almost seems like something that the database could support... Brett