Re: Logical log write volumes

Andy Kent writes:- > > Andy, if you want to work out the contents for logical logs for a > > particualr transaction you can use tbcheck to get the details of the > > transaction. This tells you how much space in the logs are used for > > each component of the transaction. > > Yes, but what I was after was a means of estimating log usage without > actually having to run it. The best method is to determine the number of rows changed by the transaction, add their sizes together, multiply by 2, and then add approx 25% to allow for headers etc. It's CRUDE but it should work. > > > As to your question concerning the > > difference between buffered and unbuffered logging, I can't believe > you > had to ask. The amount of logging does not change, merely the > frequency > > of flushing the buffer to disk. > > So what about "piggyback" writes then? Are they a figment of my > imagination? > > > akent@cix.compulink.co.uk (Andy Kent) > ------------------------------------------------ > Redland, Bristol, England Piggy-back COMMITs are a real event. All this means is that the logical log buffer in shared memory is flushed to disk periodically. If a second commit occurs before the log buffer is flushed it will continue flushing the log buffer to disk. This does not alter the amount of data written to the log, Malcolm Weallans Online Database Consultancy 2 Arkley Court Maidenhead Berks SL6 2YR Phone 0628-72154 Fax 0628-37463 CIX - onlinedbc