Re: How to best use 32 15k.7 300GB drives? - Mailing list pgsql-performance

From Віталій Тимчишин
Subject Re: How to best use 32 15k.7 300GB drives?
Date
Msg-id AANLkTinX3VHp0CQZk1T+JygMVUA7yiD9OAPbJ6wvEA7Z@mail.gmail.com
Whole thread Raw
In response to Re: How to best use 32 15k.7 300GB drives?  (Robert Haas <robertmhaas@gmail.com>)
List pgsql-performance


2011/2/4 Robert Haas <robertmhaas@gmail.com>
On Fri, Feb 4, 2011 at 4:19 AM, Vitalii Tymchyshyn <tivv00@gmail.com> wrote:
> Why do you expect such a invasive code changes? I know little about
> postgresql code layering, but what I propose (with changing delete to
> truncate) is:
> 1) Leave tuple addressing as it is now

i.e. a block number and a slot position within the block?

Seems like you'd need <file,block,slot>.

No, that's what I mean. Leave as it is. You will have file logical length (fixed for all but the last one, 1GB currently) and file actual legth that can be less (if file trucated). In the latter case you still have this "empty" blocks that don't exists at all. Actually the simplest implementation could be to tell to file system "drop this part of file and pretend it's all zeros", but I don't think many FSs (OSes?) supports this.
So, each  file still have it's fixed N blocks. And filenumber is still blocknumber / N.


--
Best regards,
 Vitalii Tymchyshyn

pgsql-performance by date:

Previous
From: Greg Smith
Date:
Subject: Re: [HACKERS] Slow count(*) again...
Next
From: "Pierre C"
Date:
Subject: Re: Really really slow select count(*)