Re: Large XML Data, Parsing, JDBC, RMI, 2 Oracle DB

I don't have a solution for you, the problem is too complex with the basic description you provided. But it does sound like the Parser is the bottleneck. Java by default uses "Green" threads, meaning basically that the JVM knows how to schedule processes and the OS never gets involved. This keeps the necessity for knowing a lot about all the different OS's that Java can run under out of the picture. You can however utilize "Native" threads in Solaris and at least kind of under Windows. I have never had to implement this so I am now in "guessing" territory and will conclude with a recommendation you pick up a copy of Java Threads from O"Reilly by Oaks & Wong. They go into a lot of detail about supporting multiple CPU's in the Solaris/Unix environment and it may have at least part of the answer you are looking for.

HTH, Greg Sliker

"Raden Java" <radenjava_at_theglobe.com.invalide> wrote in message news:tlilner8k3pd5d_at_corp.supernews.com...
> Hi,
>
> I'm trying to insert a large of data over 25,000,000 elements in xml
 format to
> 2 different Oracle Database on 2 different server.
>
> I'm currently working on 4 machines, machine A 1 and 2 is a database
 running
> Oracle 9i, machine B is a parsing machine (8 CPU Solaris), and
> machine C is where the RMI Connection to Database.
>
> What, I'm doing right now is I'm parsing every elements from the files
> using Oracle Parser and insert it to database using jdbc.
> But the program isn;t using all of the cpus because
> it's only capable to Parse and insert around 18,000 elements /
> hour, 432,000 elements in 1 day.
> So to parse all of data it will take around 57 days ?
> It should be something wrong here, but I don't have any clue on how to
> get through it.
>
> My only thought is to modified the parsing program, to utilize the 8 cpus
> resources.
>
> Can someone guide me how to find resource in Parallelism, so the Parsing
> program will utilize the SMP.
>
> Or if anyone have some advice on the best way to implement it, I really
> appreciate it.
>
> Thanks
Received on Sat Jul 21 2001 - 09:01:48 CDT