| From: | Gavin Sherry <swm(at)linuxworld(dot)com(dot)au> | 
|---|---|
| To: | Curt Sampson <cjs(at)cynic(dot)net> | 
| Cc: | pgsql-hackers(at)postgresql(dot)org | 
| Subject: | Re: Importing Large Amounts of Data | 
| Date: | 2002-04-16 05:17:25 | 
| Message-ID: | Pine.LNX.4.21.0204161511130.8569-100000@linuxworld.com.au | 
| Views: | Whole Thread | Raw Message | Download mbox | Resend email | 
| Thread: | |
| Lists: | pgsql-hackers | 
On Tue, 16 Apr 2002, Curt Sampson wrote:
> > Given the very low parsing and 'planning' overhead, the real cost would be
> > WAL (the bootstrapper could fail and render the database unusable) and the
> > subsequent updating of on-disk relations.
> 
> MS SQL Server, when doing a BULK INSERT or BCP, can do it as a fully or
> "minimally" logged operation. When minimally logged, there's no ability
> to roll-forward or recover inserted data, just the ability to go back
> to the state at the beginning of the operation. This technique can work
> even though an on-line database. A bit more information is available at
The other reason I say that this bootstrap tool would still use WAL is
that bypassing WAL would require writing a fairly significant amount of
code (unless the pre-WAL heap_insert() code could be used, with relevant
modification).
On the other hand, I would imagine it to be very difficult to implement
an 'interactive' roll back facility with the kind of tool I am
describing.
Gavin
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Hiroshi Inoue | 2002-04-16 05:23:48 | Re: ANSI Compliant Inserts | 
| Previous Message | Bruce Momjian | 2002-04-16 05:15:11 | Re: [PATCHES] YADP - Yet another Dependency Patch |