WP_Data Management Best Practices

43 / 96

W H I T E P A P E R
© 2017 Persistent Systems Ltd. All rights reserved. 43
www.persistent.com
— The alpha period is the data warehouse team’s first opportunity to conduct an end-to-end system test before
going live containing the following items:
— Hardware and software installation including initial systemand user configuration.
— An automated ETL system test suite composed of test scripts containing (i) data extraction against a
known, static set of data (containing both correct and imperfect data), (ii) initial loading tests, (iii)
incremental data processing tests, and (iv) queries contained in predefined reports. Build an
application, the regression test suite, which compares and logs the result of all the runs from the
test suite against known correct results. Make sure you run it always before releasing your ETL and BI
system.
— Data quality assurance testing. This is described in section below.
— Operations process testing. This consists of the following steps.
— Verify end-to-end operations first using the static regression testing dataset. Make sure that the
imperfect data generates predictable exceptions and associated management. For instance, when
receiving a late arriving dimension, test the handling of SCDs associated with such late arriving
dimensions, recalculations, and re-aggregations; furthermore, make sure these don’t halt processing.
— Verify that jobs start when they should, that they run correctly under normal operation, and that
standard reports are built on schedule.
— Then, move on to alternative datasets allowing testing for handling unusual events (e.g., receiving
data that is garbled, or has unbelievably high row counts). Test the procedures handling these events
–this includes invoking database tasks such as table space management, as discussed in the section
above.
— Finally, move on to injecting unpredictable errors such as a network failure or a power outage, and test
that the restart capabilities of the systemwork as they should.
— Live Testing. Once the historical data is accurate and the operational processes are clean, real data
testing may start. In the test environment, point to the same operational sources you will use in production.
Run the live test long enough so that you can see real world patterns in the data and operational
environment. If there is an operational cycle that may affect the system, such as a monthly closing
process, make sure the tests span a complete cycle.
— Performance testing. This is described in section below.
— In alpha-testing generally no access to business users is given. This changes in beta testing, where the main
goals are (i) to conduct an end-to-end user test going through the same items of alpha testing and adding
items such as end user application quality, training materials, support infrastructure and user
communication, and (ii) to work out any remaining glitches. The data warehouse team should also use the
beta period to verify the necessary back-room operational processes before the user population expands,
including back-up and archiving procedures, disaster recovery, stress testing, and performance monitoring,
as described in chapter 13 o