Mailing List Archive


Back to the month index Back to the list index
Jean-Olivier Holingue (joh@shore.net)
Fri, 21 Nov 1997 11:41:45 -0500 (EST) 

Date: Fri, 21 Nov 1997 11:41:45 -0500 (EST)
From: Jean-Olivier Holingue <joh@shore.net>
Subject: Re: S.O.S. H~~E~~L~~P!!!
Message-ID: <Pine.SUN.3.96.971121113250.28028A-100000@northshore.shore.net>

I don't know what is the magic formula to compute the size of an index
(there was one with v1.x for primary key, but I have not seen anything
regarding v2.0.x)
Anyway, what you see is the "reserved' space of the index; in other words,
if you are adding more rows to you table, your index is not going to grow
anymore until the size of the index (based on the lenght of the indexed
field) will go beyond the original 1 meg limit. You could test it very
easily by importing more records in your datafile. I did dump hundreds of
records into your example and the index stayed the same.

Now, if you plan to have 300Mb of raw data and don't want to use more than
500Mb of disk storage, you are really asking for troubles :> using 900
tables to improve performances is also a little extreme :> :>

Guess you should go back to the drawing board or buy an extra disk for
your client...

HTH

Jean-Olivier Holingue
joh@shore.net

On Fri, 21 Nov 1997 yuq-msql@wayx.com wrote:

> hi,all
>
> I need your HELP.Please DO give some attention,Especially those MiniSQL
> experts and guys from Hughes.
>
> My question is about the index of minisql 2.02.I found the index file
> size is something fixed.What I mean is no matter how many rows I have in a
> table,the index file's size is the same.
>
> I have such a table:
> create table table1
> (
> yunit char(38),
> yleader char(20),
> ypostcode char(6),
> ytelcode char(5),
> ytelnumber char(14),
> yaddress char(58),
> yservice char(20),
> yclasscode char(4),
> yareacode char(6),
> yusercode char(7),
> ypasswd char(8),
> yexpire char(10),
> yreserve1 char(4),
> yreserve2 char(4)
> )
>
> and create a index by such statement:
> create index idx_yunit on table1 ;
>
> After inserting only ONE record into the table by msqlimport,I got the
> following list by ls -l:
>
> -rw------- 1 nobody other 14369 Nov 21 20:57 table1.dat
> -rw------- 1 nobody other 2184 Nov 21 20:34 table1.def
> -rw------- 1 nobody other 288 Nov 21 20:58 table1.idx
> -rw------- 1 nobody other 1146912 Nov 21 20:58 table1.idx-
> ^^^^^^^
> idx_yunit
> -rw------- 1 nobody other 4 Nov 21 20:34 table1.ofl
>
>
> WHY?Why the idex file size is so large?I only have one record at all.
>
> May be you will say that the strategy of Hughes.But now,I have about 991
> tables with same structure as above.I divide them into so many tables only
> for geting faster query speed.The total data are may be 300MB,I think I can
> hold all the tables within 500MB,but if a table with only one record will
> have a 1MB index file,how about 991 tables?
>
> I have promise my customer that I can hold them in about 500MB.What should
> I do now?Should I run query without index?
> Any one can give me some solutions or explanations?Are there some one from
> Hughes?
>
> I don't know what will happend if I can not hold all tables in about
> 500MB.I don't know what my customer will do.So,Please Do help me.
>
> Thank you a lot.
> any help plese mail to > any help plese mail to yuq-msql@wayx.com.
>