Thanks Jose,

You've answered my questions pretty well.

Rod


On 8 February 2012 20:51, Jose Huerta <[email protected]> wrote:

> ** Indexes in remedy are mainly text indexes. Normally other applications
> index keys (primary and foreign) and search fields. Since in Remedy keys
> are text based, almost all indexes are text based.
> With my script I duplicate only text based indexes (no date or number).
>  But this means almost all indexes.
>
> I recommend to not delete case sensitive indexes by three reasons:
>
> 1.- Indexes are maintained by ARS. I always try to not touch the ARS
> managed objects, since it is unsupported.
> 2.- I maintain an easy rollback by turning off the login trigger. It is
> unsupported, so it can come the day when something goes wrong and it is
> beacuse of the case insensitive mode.
> 3.- Some apps (like SQL Developer, BI, or others) can connect to the
> database and force a sort-where mode. If they force the case sensitive
> mode, all case insensitive indexes will be useless.
>
> Since we have a lot of indexes in Remedy and almost all are duplicated
> this means an increase in database size.
>
> Non indexed queries run slower. I think that this is because the system
> must apply the conversion function to all registers at the table to perform
> the comparison. And this is an extra load. The fact is that if a search
> cost 10 seconds, with case insensitive costs 12 seconds. But this applies
> only to non-indexed. Our DBA scans non-indexes queries to look for a
> pattern and propose new indexes.
>
> Finally, the oracle index corruption statement: I don't know if index are
> corrupted or not. But we experience a problem: After some months, Oracle
> don't use case insensitive indexes and starts performing full text scans.
> Don't know why. The recommendation from one DBA was to drop and create all
> indexes. He talked about "index corruption". I don't know if indexes where
> corrupted or not, but dropping and creating them again solved the problem.
> So our DBA has a programmed maintenance task of dropping and creating all
> indexes each two months.
>
> Regards,
>
>
> Jose M. Huerta
> Project Manager**
>
> Movil: 661 665 088
>
> Telf.: 971 75 03 24****
>
> Fax: 971 75 07 94****
>
>  <http://www.sm2baleares.es/>****
>
> SM2 Baleares S.A.
> C/Rita Levi ****
>
> Edificio SM2 Parc Bit****
>
> 07121 Palma de Mallorca****
>
>           <http://es-es.facebook.com/pages/SM2-Baleares/158608627954>    
> <http://twitter.com/#%21/SM2Baleares>
>      <http://www.linkedin.com/company/sm2-baleares>
>
> La información contenida en este mensaje de correo electrónico es
> confidencial. La misma, es enviada con la intención de que únicamente sea
> leída por la persona(s) a la(s) que va dirigida. El acceso a este mensaje
> por otras personas no está autorizado, por lo que en tal caso, le rogamos
> que nos lo comunique por la misma vía, se abstenga de realizar copias del
> mensaje o remitirlo o entregarlo a otra persona y proceda a borrarlo de
> inmediato.****
>
> P Por favor, no imprima este mensaje ni sus documentos adjuntos si no es
> necesario.
>
>
>
> On Wed, Feb 8, 2012 at 13:30, Rod Harris <[email protected]> wrote:
>
>> ** Hi Jose,
>>
>>
>> You are right in saying that Remedy defaults to case sensitive indexes
>> for Oracle and to avoid performance issues when case insensitive you must
>> maintain case insensitive indexes outside of Dev Studio/Remedy.
>>
>> Likewise I agree that performance on indexed queries will be
>> indistinguishable between the two sensitive/insensitive if the index is
>> used.
>>
>> I'm interested in the reasons why you're quoting a large increase in
>> database size and decrease in performance on non indexed queries. Is it
>> because in your technique you are recommending duplicating indexes, one for
>> each mode?
>>
>> I'm also interested in your statement that the insensitive indexes become
>> corrupt over time.
>>
>> For Remedy it only does inserts/deletes one database row at a time so it
>> is difficult to see any performance differences in these operations so it
>> should rarely be a consideration.
>>
>> Rod Harris
>>
>>

_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org
attend wwrug12 www.wwrug12.com ARSList: "Where the Answers Are"

Reply via email to