hatman@vienna weblog

So finally I migrated public blog

Hope I will find time soon for posting…

Thanks to Damir aka “Funky” from Zagreb to ask me for this blog!

As I said him, I’m trying to get space for one XE instance on provider server…

In XE sholud be enough space for my new blog
… something in style of AskTom ))
…consulting related….

Begining of this August was very intensive and I found myself in position of having no time for free activities like this blog.

Anyway lot of work is the best way to prepare myself for meeting with the Notre Dame in next weeks.

It’s been a busy week. I didn’t have free time to work with WebCenter seriously, but soon is vacation ….

Speaking of which, we are leaving for France in second half of August. We’ll be there for about two weeks and we searching at the moment a home-swap. A family from Paris can stay in our home in Vienna. If we can wish a place then that will be Fontenay suis Bois, a suburb of Paris, and very close to some friends, which will be nice.

Well, if some of you are interested for home-swap (Vienna-Paris), please contact me by email.

Well, we have here in Vienna fourth day 39°C (102°F) temperature.

Only place where was/is possible to feel fine is if you are near to the river Danube. Not that I don’t like air-conditioned rooms, but I’m doing this blog freebie, so I will always avoid such activity when outside is such a weather that you have something like Santa Monica Beach in the Middle of Europe …

Some of you asked when will appear 11g?

Well, it’s almost happen! Oracle Database 11g Launch is on Wednesday, 11th July at 16.00 h SWET.

You can watch a live this happening here:

http://www.oracle.com/pls/ebn/live_viewer.main?p_direct=yes&p_shows_id=5717957

* SWET = Summer Western European Time (GMT + 2 hours)

One of most interesting person from Oracle Corporation at the Conference, I found Bryn Llewellyn and his session about PL/SQL performance.
Bryn compared in tests Oracle 9i R2 and 10g R1. Oracle Database 10g brings a brand-new optimizing PL/SQL compiler and a newly tuned PL/SQL execution environment. Result is dramatic performance improvements.
PL/SQL computational programs should run at least twice as fast as they did in Oracle9i Database Release 2.
This, in turn, is about 1.5 times as fast as Oracle8 Database!

The optimizing compiler changes the way programmers should think about traditional hand optimizations. The significantly improved PVM increases the speed of underlying operations
like the addressing of collection elements and eliminates some historical asymmetries…

Bryn expalind to us in detail his “Clean Room” Experiment. In short:
1. Startup the database on otherwise quiet machine
2. Run the test once to warm up
3. Run it again and time it (Dbms_Utility.Get_Time)
4. Shutdown the database
5. Record the results elsewhere for later analysis
6. Time each test many times

The most important advices for PL/SQL developers regarding results of test are:
1. Now you needn’t care for expression inside a loop if it remains constant over all iterations of the loop
2. Using CONSTANT in 10g helps performance, so be careful to use CONSTANT when you can
3. There’s no significant performance penalty if you break a complicated big expression into pieces for readability
4. There’s no performance penalty if you declare variables, types and procedures deep inside nested loops for readability
5. Use hand-optimization to avoid referring to a collection element within a loop when its address remains constant over all iterations is no more good advice.
6. Regarding Callbacks: Sometimes, when you write “library” code for re-use, you need it to be able to call a procedure with a fixed parameter signature that your customer will write. Since you don’t know its name when you compile your code, you have to use the usual approach for such situations: dynamic SQL. However, because this is slow and because there’s a risk of runtime error that you’d prefer to catch at compile time, you should try to find a radically different approach to give your customer the required extensibility hook. So we have to use dynamic SQL to implement a callback.
Dynamic polymorphism, a.k.a. virtual procedure call, is the natural way to implement a callback. It’s quick and it also allows compile time checking and the normal benefits of cascading invalidation. You absolutely do not have to usedynamic SQL.
This will cost you hugely in call overhead was true before 9iR1 – things have changed: polymorphism helps performance!
7. Dynamic SQL versus static SQL:
- Dbms_Sql is never the winner. Use NDS.
- NDS single-row was slower than static single-row in 9iR2 but they’re similar in 10gR1 because of cursor reuse.
- NDS and static SQL were similar for the nonbulk case in 9iR2 but static leaps ahead in 10gR1 because the PL/SQL optimizing compiler bulkifies it.
- NDS and static SQL are similar for the bulk case
Neither static SQL nor native dynamic SQL avoid soft parse when using a REF CURSOR.
New in 10gR1 was that native dynamic SQL avoids soft parse when not using a REF CURSOR.
For single row select and for bulk syntax, native dynamic SQL is therefore now about as fast as static SQL.