Wednesday, July 18, 2012

Data Migration Testing_Staging servers

The paradigm differences in DEV and TEST staging servers.

 Test staging extracts all the data in the destination data model alignment. Whereas DEV will need a differential extraction and progressive transformation of data.
 Test staging will only bring in the records that SHOULD be in the destination. Whereas DM extraction should get in all the records from Source and log a comment on any exceptions or failure migrations.
 Test staging will be built with the primary Key/constraints based on the destination model. The repeated test efforts should be a mere LEFT/RIGHT/INNER join validations between TEST STAGING and DESTINATION. Anything missed or in excess is a bug kind of approach.
 Test Staging tables are usually build with a Second NORMAL form or Third Normal form based on the rules/ Index of migration. With an exception of DATA CONSITENCY


I. Build the test staging
Building the test environment (Test staging server)
In any project that has a data + functional changes happening in parallel the following needs to be ensured with regards to designing the test environment/data base.

 Isolating the migrated data from SIT test manipulation. Unless this is done we never know what was a Data transformed by the Data Migration VS what was a bad data caused by functional test tweaking.
 Get all the Source/Staging and Destination Databases on a single server to foster better performance in high volume data test.
 Instrument complete set of the daily jobs and other Procedures that recalculates data in the system (Desired/Undesired changes roll back).
 Staging server should be planned to accommodate a Space GROWTH by atleast 200%.
 All the test results for each refresh should be available on the test staging to see the differential data changes associated with code fixes.
Scope VS Complexity VS Refresh Ratio:
Source DB counts Number of Master DB destination Complexity of Data in terms for transformations Minimum No. of Refreshes required in SIT Suggested test span on each refresh
1 1 Apple to Apple/upgrades/SQL to Oracle data transformation 2 5 days
2 1 Less than 10 data merge/split DM rules 4 5 days
>2 1 Less than 10 data merge/split DM rules 4 5 days
>2 >1 More than 25 data merge/split/transform DM rules spanned across more than 50 tables 7 10 ays

No comments:

Post a Comment