Re: Duplicate databases

->From: jh@cadre.com (Joe Hartley) ->Subject: Duplicate databases ->Date: 30 Mar 93 15:02:31 GMT ->Reply-To: jh@cadre.com ->Organization: Cadre Technologies, Inc. -> ->We've got a Sun 4/470 running SunOS 4.1.3 and an accounting/manufacturing/ ->order admin application that uses Informix/SE as its underlying RDBMS. ->The app requires SE; On-line is out of the question. -> ->The problem: At the end of the month, the orders are coming in fast and ->furious and the OA and production staffs are working full steam trying to ->get the orders out. Unfortunately, at the same time, the beancounters ->and salespeople run all sorts of CPU and I/O intensive reports trying ->to figure out where the sales are in relationship to the forecasts. ->This has the effect of bringing the app to its knees, making it more ->difficult for the OA and Prod folks to get the real work done. -> ->What we'd like to do is find some way of mirroring the main database ->in a timely fashion so that the reports can be run off of the DB ->copy, keeping the record locking to a minimum on the working database. -> ->I don't think that RAIDs are the answer here, just some sort of brilliant ->software solution. Has anyone faced and conquered this problem before? -> ->============================================================================== ->Joe Hartley | jh@cadre.com - Whenever you find that you are on the ->Cadre Technologies | side of the majority, it is time to reform. - M. Twain ->222 Richmond St. | ------------------------------------------------------- ->Providence, RI 02903 | Overman 1st Class - the Kilgore Trout Memorial Clench ->(401) 351-5950 x266 | of the Church of the SubGenius ->============================================================================== Good morning, Joe, Before you try to mirror your database, may I humbly suggest that you revisit your indexing. I was able to significantly improve the performance of one of my databases by adding an index to a table that I originally had without index. Indexing is a two-edged sword. Maintaining the indexes will slow your insert and update operations, i.e., order writing. But indexes may speed your reporting enough to offset the load and then some. Another option may be to have a middle of the night cron job run about the 25th of the month to build indexes, and have another such job drop the indexes on the 3rd or so. Thus you could run 2/3 of the time without indexes, and only put up with them when the bean counters are reporting. Richard Ridley's suggestion of using 'dirty read' isolation level for reporting is a good one, but I believe it only applies to Online, not to SE. Regards, Alan +------------------------------+---------------------------------------+ | R. Alan Popiel | Internet: alan@den.mmc.com | | Martin Marietta, LSC | ( Please note: My opinions do not ) | | P.O. Box 179, M/S 5422 | ( represent official Martin policy. ) | | Denver, Colorado 80201-0179 | Voice: 303-977-9998 | +------------------------------+---------------------------------------+