|
From: Vlad K. <hv...@us...> - 2010-01-08 13:47:56
|
>> > Also please pay attention - the fact that server does not reuse pages
>> > after 'DROP TABLE' does not explain performance degradation. If for any
>> > run each table containt X pages, why should we have visible performance
>> > difference when this pages are located in the beginning of the file or in
>> > the end of it. Specially well seen in series 11-20 from original Gary's
>> > letter.
>>
>> Seems like DROP and subsequent CREATE TABLE was failed. And
>> garbage collection is blocked during test run.
>
> But why did database become smaller after backup+restore in that case?
> Remember, no DELETE FROM, only DROP.
Good question ;)
We can guess a long time without more info. I would like to see gstat -r
output during test run (as Ann already told). Also it is interesting to see
results of gfix -v -full after the runs.
Regards,
Vlad
|