On Tue, Feb 28, 2006 at 11:58:44AM -0500, Bruce Momjian wrote:
> Jim C. Nasby wrote:
> > On Tue, Feb 28, 2006 at 01:18:14AM -0500, Greg Stark wrote:
> > > But I think the thought process went the other direction. If you have the bit
> > > intended for index scans indicating that the tuple is not "in doubt" ie, it's
> > > visible to every transaction, then that also implies the tuple doesn't need to
> > > be visited by vacuum.
> > >
> > > Skipping pages that don't contain any in doubt tuples would be a huge win.
> > > Even if there might be some additional pages that vacuum could in theory be
> > > skipping too.
> > Agreed. IMO, *anything* that improves the efficiency of vacuum would be
> > of huge benefit, and keeping a bitmap of pages that are known to be 100%
> > visible would be a big start in that direction. For many large tables,
> > this case would cover a large percentage of pages, speeding up not only
> > vacuum but also index scans.
> > ISTM that the continuing debate about how to improve vacuum is due
> > largely to the fact that there are a very large number of possibilities.
> > I would very much like to see a decision on one to impliment as a
> > starting point. Ideas about some kind of dead-space-map, or a
> > known-clean-map have been floating around for at least 2 versions now.
> Right, we should discuss all the possibilities and do something. I
> think we just don't know what yet.
I guess that might make more sense if the discussions were more
organized so that we could go back later and easily find the pros and
cons of all the options, but that doesn't really exist. Sure, there's
the archives, but trying to read through discussions there is a PITA,
let alone trying to find all the relevant bits in the first place. I
doubt that will change without bringing some kind of more formal
collaboration tool online (something I really doubt folks would go for).
In this case, ISTM that most of the ideas have already been pretty well
hashed out, and I can't think of anything that would be a better place
to start than either a known clean map or a dead space map. The
advantages of each seem pretty clear and I think the implimentation
issues are pretty well hashed out.
Jim C. Nasby, Sr. Engineering Consultant jnasby(at)pervasive(dot)com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
In response to
pgsql-hackers by date
|Next:||From: Andrew - Supernews||Date: 2006-02-28 23:25:09|
|Subject: Re: bug in PG_VERSION_NUM patch|
|Previous:||From: Lukas Smith||Date: 2006-02-28 23:02:55|
|Subject: Re: [PERFORM] temporary indexes|