I installed Oracle10g recently on a test server. I expected performance and speed degradation, but not to the extent of 10% when compared to Tbase. All these years I had kept Tbase alive by keeping it trim and healthy by so many archiving routines developed by me. But this year It's rather difficult to continue with Tbase. I shall be divorcing it this year end. But the question is baan has not come out with best parameters for Oracle10g. Is anyone using Oracle 10g on Solaris 2.9 on 4 proc SUN Enterprise Server. Can I have best set of parameters which will give better performance.
Thanks in advance.

Regards

Fouzan Majid

Hi There,

Your wording is not quite clear: Did your performance became 10% slower or went down to 10% (meaning a factor of 10 times)?

10% slower is actually very good, if you are running on the same box. Oracle has much higher memory and CPU requirements. This is absolutely justified, considering the extra work oracle is putting for transaction consistency, recovery and all sorts of other topics.

Regards

Markus

Hi Mark
Sorry for that, It is reduced to 10% A session which took 2 minutes is taking more than 20 minutes. for sure I expected 50% degradation in performance. but the performance is really 10% of what it used to be.

Maybe not the best 10g parameters are available on support, but if you install it as you should do with 9i I think that performance will be nearly at optimum.

As it look likes you have nearly any Oracle experiense I suggest you to consult someone with lots of this knowledge. There is so much more than only setting a bunch op parameters.

Kind regards,
Dick

Dear Dick

I had started thinking of opting for 9i insted of 10g. Thanks for your advice. Please share some points on the performance, other than just the parameters if you can.
Thanks in advance


Regards


Fouzan

Here some ideas:
- Use Locally Managed tablespaces
- Spread the data across sufficient disks. Disk space is one thing, disk IO is more important. Put the redo on the fastest disk(s) and put the other stuff on other disks.
- Make sure you have sufficient memory (allocated) fro shared_pool and db_cache_size
- Use dbms_stats to analyze the tables
- Use statspack to check the sanity of the environment

I think this is sufficient to let you go through documentation for the rest of this year.

Kind regards,
Dick

In most cases, test servers have limited resources.
A full-blown Oracle install will require (again in most cases) a handful of CPU's, a shared memory pool the size of Texas and, as Dick already mentioned, a couple of weeks worth of tweaking and twiddling.

Make sure you have enough resources available on whatever box you run your Oracle and go two-tier if you have to.

Francesco,

Maybe to explain a bit more why a test server should be more or less the same as a production server - or at least has sufficient resources -: If you do a test on a very small system ands you face performance problems, you will not be sure if the same kind of problems will occur on the production server or it is just because the system is not setup well.

Kind regards,
Dick