TOSCA@data
2 Current situation
Efficient and effective testing requires that test cases be derived systematically. By applying proven techniques
(equivalence classes, boundary analysis, combinatorial analysis), very accurate test case designs can be determined,
so that high test coverage is achieved with a minimum of test cases.
To obtain the expected results when executing such highly specific test cases, the test data must be planned down to
the smallest detail - complex systems require precise, accurately defined initial conditions in the data object set-up in
order to execute test cases successfully.
2.1 Synthetic test data
The greatest possible control over the test data stock can be achieved if the data is generated during the test process
itself. This is what we call synthetic test data. This approach has the following advantages:
Perfect fit
The test data is determined by a structured, systematic test case design process - which we call TestCase-Design
- together with the actual test cases. Each data object has a specific test purpose, or is the result of a specific test
process - the test inputs and outputs. TestCase-Design allows the level of accuracy of the test profile to be
specified.
Synthetic test data enables great test depth by permitting complex tests of particular functionality, even when
those tests require specific histories for test data objects.
Low data volume and maintenance costs
In synthetic test data, data objects are only created if they serve a specific purpose. Therefore, there is typically far
less synthetic test data than production data; reductions by up to three orders of magnitude can be achieved.
As is the case with production data, the maintenance of test data is an ongoing task during daily post-processing,
otherwise it would be unusable. In the mainframe environment, such post-processing often requires significant
(and costly) CPU time - with synthetic test data, the CPU consumption and associated costs are minimized.
The right test data on demand
If test data is generated automatically, it is available in the desired quantity at any time, and can be extended as
required.
No sanitization required
Synthetic test data is artificial and constructed without reference to confidential customer information, therefore it
does not need to be sanitized.
Synthetic test data has a number of obvious advantages, so why is it not used more often?
When test experts try to synthesize test data, they are faced with the following obstacles:
Test execution may consume test data
Test data may be effectively consumed by test execution if the data objects are modified in such a way that other
test executions cannot use the same objects.
For example, when a valid contract is canceled, it will be marked as canceled. The same contract cannot be used
again for further test executions (unless another test were fortuitously to require a canceled contract).
Depending on the task, a large number of test data objects, in various states, may be required, and, in general, it
is not feasible to select or generate these once, and reuse them repeatedly. Consequently, test cases need to be
3星 · 超过75%的资源 需积分: 5 187 浏览量
2012-02-16
23:05:49
上传
评论
收藏 1.06MB PDF 举报 
lphua1232014-04-22内容一般般，看了帮助不大
