Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread Clemens Ladisch
O'Toole, Eamonn wrote: >> From: Clemens Ladisch >> Is it the DELETE or the INSERT that is slow, or both? > > Good question - I don't know > >> Are you addressing the records by ROWID or by name? > > By name So when the DELETE searches for the record to be deleted, it has to scan the *entire*

Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread O'Toole, Eamonn
Sqlite_analyzer output: /** Disk-Space Utilization Report For container10M.db *** As of 2012-Nov-09 13:04:53 Page size in bytes 1024 Pages in the whole file (measured) 3412536 Pages in the whole file (calculated).. 3412535 Pages that store

Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread O'Toole, Eamonn
Thanks Clemens, comments below: > -Original Message- > From: sqlite-users-boun...@sqlite.org [mailto:sqlite-users- > boun...@sqlite.org] On Behalf Of Clemens Ladisch > Sent: 09 November 2012 12:26 > To: sqlite-users@sqlite.org > Subject: Re: [sqlite] Anomalously slow performance on

Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread O'Toole, Eamonn
Thanks Michael. Comments below: > -Original Message- > From: sqlite-users-boun...@sqlite.org [mailto:sqlite-users- > boun...@sqlite.org] On Behalf Of Black, Michael (IS) > Sent: 09 November 2012 13:49 > To: General Discussion of SQLite Database > Subject: Re: [sqlite] Anomalously slow

Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread Black, Michael (IS)
Hmmm...is this a disk head seeking problem? You've got several TEXT entries which are either NULL or default to ''; I did a small experiment: on 3.7.13 CREATE TABLE container_stat ( account TEXT, container TEXT, created_at TEXT,

Re: [sqlite] Anomalously slow performance on updates to early entries in a DB

2012-11-09 Thread Clemens Ladisch
O'Toole, Eamonn wrote: > We are seeing a performance anomaly on updates to existing object > records in the SQLite DB. If the container DB is sufficiently large > (about 10 million objects, 3.3GB) then the time to update records at > the beginning of the database by order of entry is anomalously