maxdb on sparc server

[Former Member]

hi

i have an instance of Maxdb version 7.6.05.09 running on sun solaris X86.

i make a migration work to Maxdb version 7.6.05.09 over Sun solaris Sparc server, make an export of all productive schemas with data; import schemas ,data and run correctly, but when i make the same query in both databases, in the sparc database take more time than x86 server.

the query is that

SELECT

MAX(SCDI_FECHASALDO) , CUEN_ID

FROM TRANSACCION.SALDOCUENTADIARIO

GROUP BY CUEN_ID

the response time on X86 server is

Statement 'SELECT MAX(SCDI_FECHASALDO) , CUEN_ID FROM TRANSACCION.SALDOCUENTADIARIO GROUP BY CUEN_ID' successfully executed in 328 ms.

and the response in sparc server

Statement 'SELECT MAX(SCDI_FECHASALDO) , CUEN_ID FROM TRANSACCION.SALDOCUENTADIARIO GROUP BY CUEN_ID' successfully executed in 6.156 seconds.

index on tables are the same in both servers.

how can i do in this case ¡?

thanks for yuor help.

[markus_doehr2]

Did you update the statistics of the database tables after the "migration" on the SPARC system?

Markus

[Former Member]

hi ...

yes, i updated the statistics once finished the schemas imports, rebuild all table indexes and foreign keys.

but the result is the same. the time of response is very slow.

[lbreddemann]

Hi Alexander,

is the query based on a join-view or just a base table?

If this query is just accessing a single table, new statistics won't change a thing - they are not used by the single table optimizer.

The first step to analyze this issue would be: compare the exection plans for the query on both servers.

Are they the same?

Start DBAnalyzer and enable time measurement (x_cons <dbsid> time enable) to check whether I/O times are ok on the new machine.

Also, make sure that the database parameter setup is ok.

Use the automatic parameter check script for the DBAnalyzer (if you're a SAP customer) .

Check the Wiki-Entry on that topic: [MaxDB Parameter Recommendations|https://wiki.sdn.sap.com/wiki/x/mYfzAQ].

If I should make a guess here, I'd say that the query on the old system was completely processed in cache while the new system had to get data from the disks as well. Have you checked the execution times if you run the query a few times in a row?

regards,

Lars