Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.tools -> Re: Quick question about table partitioning?

Re: Quick question about table partitioning?

From: Kev.- <java2e_at_yahoo.com>
Date: Thu, 19 Apr 2001 00:11:32 GMT
Message-ID: <3ade2d07.217792458@ns1.nothingbutnet.net>

I agree about the windows thing but I dont have the money at this time to buy a solaris box. Now I could move it over to Solaris on Intel. Would that be any good?

Thanks,

Kev.-

On Tue, 17 Apr 2001 22:18:33 -0700, "Daniel A. Morgan" <dmorgan_at_exesolutions.com> wrote:

>Personally I don't think partitioning is going to give you much bang for the
>buck though others that have worked with something similar to what you have
>may disagree ... and if they have an environment more like yours listen to
>them ... not me.
>
>In your situation I would absolutely get the thing off of anything named
>Windows. Solaris is a good choice. I would go for a minimum Enterprise 450
>with 4 CPUs and at least 2 GB RAM. By logically mapping your disk array of 9
>GB or 18 GB drives to u03, u04, etc. you will effectively partition your
>table across multiple physical drives without having to do any real mapping.
>
>With respect to your bitmapped index the two criteria are volatility (how
>often is the index modified) and cardinality. If you have high cardinality,
>lots of different values, a bitmap index is a bad idea. If this is the case
>go to BTree indexes.
>
>I will be watching to see what others suggest.
>
>Daniel A. Morgan
>
>
>
>
>"Kev.-" wrote:
>
>> Hello Daniel,
>>
>> Right now the database is about 1 million rows about 2 gigs in size.
>> It is an NT server for now but will be changing to Solaris in the
>> furture. The system has almost a gig a ram. The database is pretty
>> much reads. Very few writes.
>>
>> The user base at this time is unknown but say for starters about
>> 25-50.
>>
>> The table is structure is 4 numeric fields for lookups, one date
>> field, a varchar2 500, 250, 250 and one CLOB to hold the articles.
>>
>> The search can be either on the articles or on the varchar2 500 field
>> with a numeric field id to help filter the selection.
>>
>> I had the CLOB in a seperate table but the perfrmance when I joined to
>> it was horrible so I brought it over a one table design hopeing it
>> would help increase performance. It is a little better but still
>> slow. I also tried a bitmap index on the filter field but not much
>> improvement. I am using intermedia to search the CLOB and the numeric
>> field at the same time for narrowing down the results.
>>
>> Thanks,
>>
>> Kev.-
>>
>> On Mon, 16 Apr 2001 20:35:31 -0700, "Daniel A. Morgan"
>> <dmorgan_at_exesolutions.com> wrote:
>>
>> >The answer is far more complex than the question. What operating system
>> >on what platform with how much RAM? How many users are simultaneously
>> >accessing the table, what SQL is being run against it? What indexes are
>> >on it? And other than the CLOB field, how big is the row?
>> >
>> >Daniel A. Morgan
>> >
>> >
>> >
>> >"Kev.-" wrote:
>> >
>> >> Hello,
>> >>
>> >> The table I have will become very large over time 50+ million rows.
>> >> Would I be better off partitioning the table to get better
>> >> performance? If I do this how much speed can I gain from it?
>> >>
>> >> Also I know I have to re-create the table but can I do an insert into
>> >> select * from x even though there is a CLOB column invloved?
>> >>
>> >> I figure if I have the partitions on the category ids this would help
>> >> the query as it would only have to search on that category within the
>> >> partition. but how would it work if I wanted to search across
>> >> multiple categories at one time?
>> >>
>> >> Thanks,
>> >>
>> >> Kev.-
>> >
>
Received on Wed Apr 18 2001 - 19:11:32 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US