2. Background
Databases are growing exponentially
Terabyte sized databases are starting to be the rule rather than the exception
New hardware is often needed to ensure unhindered growth of these databases
Migrating these huge databases requires maintenance windows (downtime) that are too long for most
businesses
Businesses tend to limp along with their current hardware, avoiding the migrations
2
3. The Challenge
Finding a migration method to minimize downtime
Develop a methodology to minimize the risk to the business
Use conventional, documented and certified migration methods
Ensure data integrity and quality
3
4. The Solution
HOME: Hybrid Oracle Migration Engine
The tool uses a clever combination of multiple standard migration methods
Methodology used minimizes impact on production systems
Multiple test runs ensure optimal migration throughput
Byte-for-byte sanity checks ensure data integrity
Downtime only needed during final migration
4
5. Migration Options
Standard options:
Multi-platform migrations
Completely refreshed statistics, ensuring optimal performance
On-the-fly database reorganization during the migration
Reorganization of database layout possible (and recommended) during migration
Optional:
Database version upgrade as part of the migration
Aggregation of old data into summarized database entries
Removal of redundant/superfluous data
Upgrade of application software during the migration project
5
6. Migration Strategy
Strategy developed to minimize the impact on production: downtime only required during the final migration
Involves the use of staging environments, both on the source and target sides
Source staging environment should be representative, if not an exact copy, of the source production
environment
Target staging environment should be similar to the database server of the target production environment
Allows for stringent testing of the new production environment
Crash testing, and inadvertent corruption of the new environment can be rectified within hours, instead of
having to redo the complete migration
Allows for absolute optimization to further reduce final downtime requirements
During user test cycles, both staging environments are free to be used as optimization platforms
In an iterative process, the new database can be configured and optimized at each test migration run
Delivers a clean, reorganised database
Tables are repopulated, and indexes rebuild
Statistics are rebuild as part of the migration
Development environments
If development and Q A are aligned with production, only production instance will be migrated, and the
other instances will be cloned
If not, the development instances will be migrated separately
6
7. Standard Migration Plan
A standard project plan for the migration of a SAP/Oracle landscape is shown below, which will be used as
basis for creating the migration plan for each customer/application/landscape
Depending on the individual requirements of the customer, it might be necessary to have more than the
standard technical and two functional test migrations
Customer testing periods, as well as the migration times, will vary due to the complexity of interfaces and
business processes
Final migration time depends on the size of the database, and the throughput times achieved during the
technical migrations
If upgrading of the software forms part of the project, the time needed for this upgrade will also influence the
final migration time
7
8. Major benefits
Production downtime only during final migration
Minimal impact on day-to-day business
Byte-for-byte migration guarantee
Fallback scenario guaranteed with minimal efforts, achieved by zero-change approach on the old production
server
Complete rebuilding of database statistics
Reorganization of database layout possible during migration
8
9. Previous successes
Allowed Used
Copy Migration Total Actual Data Migration
Platform Operating System Database Size (GB) Method Window Downtime Migration Sanity Checks Window
*1 *2 *3 *4 *5
Oracle 9.2.0.6 50 rcp 36:00 4:30 0:50 1:50 6:20
Solaris to AIX 5.2 Oracle 9.2.0.6 250 rcp 42:00 12:10 3:30 5:30 17:40
SUN to IBM Oracle 9.2.0.7 500 rcp 48:00 4:30 *6 2:40 3:50 8:20
Oracle 9.2.0.7 1,600 mirror 36:00 15:00 9:50 36:00 51:00 *7
Solaris to AIX 5.3
Oracle 9.2.0.6 2,200 rcp 48:00 18:00 11:20 20:00 38:00
Tru 64 to AIX 5.2 Oracle 9.2.0.6 1,400 rcp 24:00 12:00 4:30 *8 12:00 18:00
Oracle 8.1.0.7 to
2,350 rcp 24:00 12:30 5:30 7:00 19:30
HP to IBM Oracle 10.2.0.2
HP-UX 11 to AIX 5.3
Oracle 9.2.0.7 to
8,150 flashcopy 48:00 38:00 11:30 6:00 *9 44:00
Oracle 10.2.0.2
*1 The method used to copy the database from the staging server to the target server
*2 The total time allowed by the customer for the complete migration
*3 The time from handover till the target system is up and running, and ready for UAT
*4 The actual time needed to migrate the data from the source to the staging server
*5 The time needed to complete the sanity checks, running parallel to the UAT
*6 This acceleration was achieved by redevelopment of the toolset, to be used for the "bigger" migrations
*7 Due to complete sanity checks during the three test migrations done, the customer went live when the last sanity check was 98% complete
*8 Extra gigabit network cards used between the staging servers allowed for better throughput during the migration
*9 Byte-for-byte comparison was done only on a subset of the data, where custom built methods were used to migrate the data
9