I appreciate the content of this well written article. I have been doing some of this myself, but have never had it described and documented. I look forward to Part 2.

Greg Bogard

Omaha NE

It is very useful article to everyone, I got lot of answers from this article, thanks in advance for next article.

thanking you

sreeni

Excellent article.  I appreciate the testing you have done to verify your theories and also the explanation behind each approcach.

Keep up the good work.

Hi Skarai, it's just based on my experience, not based on math. A table which contains 1000 to 5000 records (say 10 to 100 columns) on 100M LAN normally takes 1 to 5 seconds to download from DB2 on AS/400 using iSeries OLE DB driver or from Informix on Unix using Informix native OLE DB driver into a staging table with no index in SQL Server on Windows 2000 or 2003 server, configured with RAID 5 for MDF and RAID 1 for LDF using 15k RPM disks, running on DELL 2850 with 8 MB memory, performed using DTS or SSIS with no transformation or scripting (just plain direct column to column mapping).

Kind regards,
Vincent Rainardi

Thanks for the follow up Vincent. I understand your earlier statemet now given the added explanation/observation.  - Sassan

I don't recommend that you get in the habit of downloading entire tables.  It may only have 10,000 records today, but a year from now, how many will it have.  We had a consultant use that method on our database, and at first the download took hours (lots of tables).  After a couple of months, it was taking longer.  I then had to rework it to only take the records modified or added that day.  The download not only took less time, but the incremental increase was no longer there.

My recommendation:  avoid the need to rewrite in the future and go with an incremental strategy.