Hello,

maybe someone will be able to help me with that:

I have two different servers, "powerfull" live serwer IBM RS6000 and "poor" test server , Intel based Compaq with SCO Unix for some developing works ; both with BaaN Base

On the test server we have developed some new sessions, dealing with new tables in a new module, and everything is running fast. But, after import DD on live server the same sessions (even on the same data) are running very slow.
I have obserwed, that problem lies in fetch time of first row in select statements. For instance I have following select

select tkikd068.* from tkikd068 for update
where tkikd068.year inrange :year.f and :year.t
selectdo ...etc , where year is third part of four-part index.
Table has about 2000000 records, 500000 of them is matching criteria.
On test environment first row is selected after 0.17 seconds. On live environmet we must wait minutes for first row to be fetched.

What can be wrong?

the reason could be the jfs of AIX. Do you actually notice during the fetch of the first row a lot of I/O on the disk particular on the transaction part of the volume group?

It's already a few year back that I have worked with AIX, but due to the transaction principle of jfs working with tbase/bisam is not that fast.

But the difference is too much.

Hope you get more response.

Rgrds,
Han

Are your index files clean? If there were a lot of delete/inserts on the tables: reorganize and test again.

It was my first idea. Already done, did not help.

Does it change when you use the between qualifier instead of inrange? And if you make it two different where clauses, > year.f and <= year.t?

Hello,

a query I posted is just an example - the problem is that every update session runs much slower than we expected - and there are lots of different quieries in those sessions. We use indices and write our componens according to official Performance Guide. So i think the problem is not in non optimized queries.
Do you know about any tbase parameter that can make all select statements to be treated like "as prepared set" (first row fetched after whole query processing) ?

I assume both boxes run tbase...
Is there a difference in parameter settings? $BSE/lib/tbase/*param