Re: Getting results from oracle query suck!

Access will always be a lot faster for something so simple. With access you are probably not communicating with another process, but rather your process is reading a disk file. With ORACLE you're process is communicating with another process, maybe using shared memory, maybe SQL*Net (depends on connect string) and crossing a process boundary is relatively expensive. There are serious benefits to this architecture, but nobody ever said it's faster when you're running everything on one machine.

Really you need to be doing array fetches. I'm not sure how smart the OCI driver is at buffering things up, so you might actually be going back to the server for each row, which means 2 context switches.

To really help you out we need to know how you're connecting (SQL*Net or bequeath) and the source for your OCI/Java program.

Cheers

MScheuter wrote:

> Thanks for the response.
>
> The query is essentially:
> select * from CName where LastName like 'Nec%'
> Again the query execution time is great, in the range of 180 ms. The time
> it takes to retrieve what I consider a small amount of data, about 1000
> rows, is terrible. Is it reasonable for oracle to take 5000 ms to send back
> to a program 1000 rows of data with no network in the picture? I think not
> since this same operation is MS Access can once again be measured in the
> hundreds of milliseconds.
>
> I tend to think that there is a gross parameter error within the database on
> NT. Our DBA is experienced with Oracle on Unix but not NT and such has not
> found anything to tune. We have been doing traces and cost tracking through
> explain plan but this has not yeilded the information that we need. I tells
> us what we already know, that the query (build of the result set) works
> great. Explain plan does not tell us what is going on when reading the
> result set. Is there additional monitoring options that can be enabled
> within Oracle?
>
> thanks again
> -mike
>
> Sybrand Bakker wrote in message
> <925837438.21998.0.muttley.d4ee154e_at_news.demon.nl>...
> >Most likely this is a sql statement tuning issue.
> >We need to have your statement and details about the tables.
> >Then: are you using array fetching. This is possible in Pro*C and without
> >doubt it is possible in OCI. Fetching a record is the costly operation
> there
> >is. Fetch your records in batches of 100 or even 1000. Also makes sure
> >Oracle traces this session. This will provide useful detail from the server
> >side. Make sure sql_trace = true and timed_statistics = true in
> >init<sid>.ora, bounce the database and see what happens
> >
> >Hth,
> >
> >Sybrand Bakker, Oracle DBA
> >
> >
> >MScheuter wrote in message <7gn1s6$rho6_at_amber.alltel.com>...
> >>I have a table that contains customer name information (about 10 fields).
> >>There are about 1.9 million rows in the table. The table is index by the
> >>last name. I am writing application code in both C++ using oracle oci and
> >>java using again the supplied oracle oci drivers (see oracle web site).
> In
> >>a test program I am timing the execution of the select statment and then
> >>timing how long it takes to 'read' (or drain) the rows of data from the
> >>database. The timing results are as follows:
> >>
> >>Rows in Result Set Query Time (ms) Read results time (ms)
> >>2824 187 11875
> >>211 156 969
> >>1330 187 4984
> >>1740 78 6062
> >>
> >>The test program was run on both a workstation using net8 and locally on
> >the
> >>server that is running the database using a local connection (no network
> >>involved) and the results were similiar.
> >>
> >>The server is a 2 way 200 mhz intell box with 512 meg ram 80 gig hd space.
> >>Can someone please explain some areas that I can look at in order to
> >improve
> >>the performance of reading the result set?
> >>
> >>Thanks
> >>-mike
> >>
> >>
> >
> >
Received on Tue May 04 1999 - 20:29:50 CDT