On 5/4/15, Richard Hipp <drh at sqlite.org> wrote: > On 5/4/15, Mayank Kumar (mayankum) <mayankum at cisco.com> wrote: >> Hi All >> I am thinking about measuring the performance of sqlite3 write >> transactions >> after lot of delete transactions have been performance but vacuum has not >> been performed versus when vacuum is performed. Wanted to get some ideas >> here on what people think theoretically should happen. I know it reclaims >> free pages, but would this give a dramatic increase in write/update >> transactions ? >> > > Many filesystems operate faster when disk access is sequential.
I failed to mention that the VACUUM command reorganizes the content of the database so that it is (mostly) sequential in the file. > > See the last paragraph at https://www.sqlite.org/draft/dbstat.html for > information on measuring how sequential your database file actually > is. > > We would very much like to hear from you if you make any actual > measurements of performance improvements after running VACUUM. So > much depends on the OS and the underlying hardware that it is > difficult to make general statements. But case studies are still > useful. > -- > D. Richard Hipp > drh at sqlite.org > -- D. Richard Hipp drh at sqlite.org
