Skip site navigation (1) Skip section navigation (2)

Re: Serious performance problem

From: "Tille, Andreas" <TilleA(at)rki(dot)de>
To: "Claus, Hermann" <ClausH(at)rki(dot)de>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: Serious performance problem
Date: 2001-10-30 10:44:16
Message-ID: Pine.LNX.4.33.0110301114580.6117-100000@wr-linux02.rki.ivbb.bund.de (view raw or flat)
Thread:
Lists: pgsql-hackers
On Mon, 29 Oct 2001, Vsevolod Lobko wrote:

> Seems that problem is very simple :))
> MSSql can do queries from indexes, without using actual table at all.
> Postgresql doesn't.
>
> So mssql avoids sequental scanning of big table, and simply does scan of
> index which is already in needed order and has very much less size.
I forewarded this information to my colleague and he replied the following
(im translating from German into English):

hc> I expected this problem.  But what is the purpose of an index: Not
hc> to look into the table itself.  Moreover this means that the expense
hc> grows linear with the table size - no good prospect at all (the
hc> good thing is it is not exponential :-)).
I have to explain that we are in the *beginning* of production process.
We expect a lot more of data.

hc> In case of real index usage the expense grows only with log(n).
hc> No matter about the better philosophy of database servers, MS-SQL-Server
hc> has a consequent index usage and so it is very fast at many queries.
hc> When performing a query to a field without index, I get a slow
hc> table scan.  This is like measuring the speed of the harddisk and
hc> the cleverness of the cache.

The consequence for my problem is now:  If it is technically possible
to implement index scans without table lookups please implement it.  If
not we just have to look for another database engine which does so,
because our applictaion really need the speed on this type of queries.
I repeat from my initial posting: The choice of the server for our
application could have importance for many projects in the field of
medicine in Germany.  I really hope that there is a reasonable solution
which perhaps could give a balance between safety and speed.  For
example I can assure in my application that the index, once created
will be valid, because I just want to read in a new set of data once
a day (from the MS-SQL Server which collects data over the day).  So
I could recreate all indices after the import and the database is
readonly until the next cron job.  Is there any chance to speed up
those applications?

Kind regards

          Andreas.


Responses

pgsql-hackers by date

Next:From: Lee KindnessDate: 2001-10-30 10:46:36
Subject: Re: Compiling on Solaris with Sun compiler
Previous:From: Jean-Michel POUREDate: 2001-10-30 10:38:48
Subject: Re: DROP/CREATE

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group