Direct path loading

[Che Ming]

I have seen your ads:

Then, I tested it with demos but I found Direct path loading is not faster 5 times than Array DML, can you tell me why? How to tune it to best performance?

[This message has been edited by mkalter (edited 08 September 2000).]

 

[Marco Kalter]

Results will of course vary for different parameters. For example, if the table has one or more indexes there will be more overhead for each inserted row, and results will be closer. The CPU, network, and disk speeds will also affect the results.

What is the performance difference between Array DML and Direct Path Loading in your situation? What is the definition of the table that is being loaded? How many records are loaded? What is the database and server configuration?

------------------
Marco Kalter
Allround Automations

[This message has been edited by mkalter (edited 08 September 2000).]

 

[Che Ming]

My Database server is 8.1.6 runned on Windows 2000 Advanced Server, I using DOA's demos to test this feature, and my file contains 10000 rows. The result is

Insert using Array takes 2.9s and
Insert using DPL takes 2.1s

BTW: Marco, how to use UBB? You have modified my first message in order to display the image correct, right, what's wrong in my message? Why my img tag doesn't work?

 

[Marco Kalter]

When I try our demo on a Pentium II 266, 256MB RAM, Windows 2000, on a local Oracle 8.1.5 database, I get the following results for a file with 10,000 records:

Array DML: 1.19 seconds
Direct Path: 0.32 seconds

It seems that you have some additional bottleneck somewhere. Are you accessing the database over a network or locally on the server?

About the picture: the tags you used contained . The UBB codes require square brackets [ ].

------------------
Marco Kalter
Allround Automations

 

[Che Ming]

Oh, yes. I access my database through network not local. Does DPL work fine only on local database?

 

[Marco Kalter]

Direct Path Loading works just fine across a network, but you will be limited by the network latency and bandwidth bottlenecks. Comparing our 2 benchmarks it would seem that it takes about 1.8 seconds to pass the data from the client to the server, and 0.3 seconds to load it into the database.

------------------
Marco Kalter
Allround Automations