Hi,
I have developped an ETL with DI11.7.2 on Windows Xp professionnal on my laptop (CoreDuo,2G of Memory,2M cache) and my ETL had made 7 or 9 min max.
Now i have deployed the developpement on an other server more performant (Solaris10,4G of Memory,Bi Proc 1,6Ghz ) but the loading had made 30 min to be complete 
Thanks in advance
Fayssal
BOCP-BOE
fayssal 
(BOB member since 2008-04-25)
Is there a difference in the connection to the source or target databases, between development and production?
dnewton 
(BOB member since 2004-01-30)
So it takes three times longer, okay?
Your laptop might be a 3.2GHz CPU which is twice as fast just looking at the clock speed, Intel is CISC vs. Sparc a RISC CPU,… So if your job consumes lots of CPU it is possible.
The Unix platforms are good in parallel processing and disk I/O, that’s why they usually get better numbers. Why don’t you run the benchmark mentioned in my Signature below to see how you compare. There are Solaris numbers as well. And if you post the results, we can discuss the findings.
Werner Daehn 
(BOB member since 2004-12-17)
As source i have flat files, but the target is a database oracle10g and it is the same thing on my laptop and on the server.
already the job server and the database are in the same machine.
fayssal (BOB member since 2008-04-25)
No my laptop CPU is 2Ghz, for the usage of CPU during running the loading in my laptop it is come to 100% but in the solaris server the al_engine and the al_job_server don’t excess the 4,7%
fayssal (BOB member since 2008-04-25)
Where is the location of the source flat file? Is it on the network or local to the job server? It should be the latter.
You can also try putting a map_operation before your target table, and setting all of the rows to Discard. This will help tell you if the bottleneck is in the loading of the target database, or someplace earlier in the process (reading the data, or transforming it in query lookups).
dnewton (BOB member since 2004-01-30)
Yes the flat files are on the same location of the job server.I indicate that the part that takes long time is loading and moving the files.
I even try to use the API BoalkLoader in the option of the target table but there no difference!
fayssal (BOB member since 2008-04-25)
Run the customer benchmark to determine what’s wrong.
Werner Daehn (BOB member since 2004-12-17)
Thanks wdaehn, But How can i Run the customer benchmark?? 
fayssal (BOB member since 2008-04-25)
dnewton (BOB member since 2004-01-30)
DI JobServer; the source and the target are on the same server:
Source DB: Oracle 10g
Source Server: Solaris10, 4G of Memory, 2Proc 1,6Ghz
DI Server: Same as above (on same machine as source db)
Network, source to DI: n/a
DI Version: 11.7.2
Network, DI to Target: n/a
Target db: Oracle 10g (on same DB as source)
Results
<DF_Benchmark_read_MS>: 300 seconds
<DF_Benchmark_API_bulkloader_MS>: 310 seconds
<DF_Benchmark_regular_load_MS>: 11814 seconds
<DF_Benchmark_single_thread_MS>: 187 seconds
<DF_Benchmark_lookup_DOP1_MS>: 135 seconds
<DF_Benchmark_lookup_DOP10_MS>: 103 seconds
What do you think?
are this results normal??
fayssal (BOB member since 2008-04-25)
(Because your tests have the_MS at the end, it looks like you did the Microsoft benchmark on your Oracle database?)
Those results look nasty. You’re saying the DF_Benchmark_regular_load_MS dataflow took almost 3 hours to run?
Something is definitely not right.
dnewton (BOB member since 2004-01-30)
It’s perhaps be the configuration of Data Base (SGA Memory, DB Writter…)!!
No??
What do you think you
fayssal (BOB member since 2008-04-25)
As said in the other thread, this line worries me most:
<DF_Benchmark_single_thread_MS>: 187 seconds
This is a row_gen -> discard test which does not include any database operation.
What was the monitor sample rate for this run? Was the system busy at that time? how much CPU did the al_engine take during then?
Werner Daehn (BOB member since 2004-12-17)
No the system was not too busy, in general for all my loading the al_engine does not go up to 25%.
fayssal (BOB member since 2008-04-25)
25%? Sounds like your 2CPU Solaris has two cores each and the DI process is all busy in one thread, the others are almost idle?
Werner Daehn (BOB member since 2004-12-17)
25% it’s for the Benchmark test, but for my developpement you can see the attachement (usage of memory and cpu).
fayssal (BOB member since 2008-04-25)