Successful Oracle to DB2 Migration at a Top 50 Financial Services Firm

Posted by Frank Fillmore on July 25, 2011 under DB2 for Linux Unix Windows, DB2 Migrations, Information on Demand Conference, InfoSphere, Oracle, Q-Replication. Tags: , , .

Whew!  The last four months have been a whirlwind.  A colleague, Jim Herrmann, and I have been working since March on data migration from Oracle to DB2 for a worldwide Top 50 integrated financial services firm.  A team of specialists from IBM and The Fillmore Group – including Joe Geller, Teresa Wan, Rebecca Bond, and John Schatko – assisted the customer with all aspects of application performance testing and tuning, code remediation, and data migration.  On Sunday, July 17 the production cutover occurred – and it was a huge success.  Props especially to IBMers Jeff Richardson, Jay Lennox, and Christian Zentgraf for providing technology leadership.

One of the keys to the migration of the data was the need to limit the site outage into a window no longer than 12 hours, from midnight to around noon on Sunday.  For an earlier migration, smaller in terms of data volume, the customer had successfully used the IBM Data Movement Tool (IDMT).  Given the metrics of the latest migration, the IDMT approach alone would have required an outage of up to 5 days – unacceptable for an online banking system on which millions of customer rely to pay bills and check balances.

The Fillmore Group proposed using InfoSphere Replication Server (Q Replication) to copy updates to the live OLTP data from Oracle to DB2 while the batch IDMT was running asynchronously to move the data from Oracle to DB2 tables.  By using IDMT and Q Replication in combination, we were able to reduce the outage to a much more manageable 12 hours.

We followed a template The Fillmore Group had used in 2010 at Constant Contact for their no-outage upgrade from DB2 for Linux, Unix, Windows v8.2 to DB2 v9.5.  Dan Berry and Sankar Padhi delivered a detailed technical overview of that project at the IBM Information On Demand conference last year: Zero Downtime Upgrades Using Q-REP at Constant Contact  The Oracle migration was incrementally more complex, but not all that difficult.

4 Comments so far

  1. Jeff Richardson July 26, 2011 3:40 pm

    Thank you for the kind words, Frank. I’d also like to thank some members from my team: Camalla Haley Baruwa, Romeo Lupascu, Peter He and Scott Martin. We put in some amazing hours and achieved amazing results. Your team was always professional and eager to take on any challenge this project presented.

  2. Roger Midgette August 8, 2011 10:30 am

    Congratulations on another successful Fillmore Group project!

  3. G August 18, 2011 11:31 am

    Curious why you did not use IBM InfoSphere Change Data Capture for Oracle and DB2 to perform this migration?

  4. Frank Fillmore August 18, 2011 12:06 pm

    Hi Glenn. The customer already had DB2 Homogeneous Replication Feature (Q Replication) installed and working in production for another database. With the existing skills base and familiarity with Q Replication, its ASNCLP scripting and monitoring, it didn’t make sense to bring in a new technology. In addition, InfoSphere Change Data Capture (ICDC) naturally writes to a target across a TCP/IP link. In this case the InfoSphere Replication Server (IRS – Q Replication for heterogeneous sources and targets) use of WebSphere MQ as a transport layer enabled us to cache the Oracle OLTP deltas in MQ while we were processing the batch IDMT migration. Finally, IBM just announced IBM InfoSphere Data Replication, an amalgam of both ICDC and Q Replication, to cut down on the “Which IBM replication technology do we use?” marketplace confusion.

Leave a Comment