Test automation and NLS in the BW 7.3 environment

To gain valuable time and minimize risk, the joint HP and DataVard team deployed a newly developed solution to automate SAP testing.

The Kate software generated 320 test cases for 110 queries and replaced manual tests. The internationally active company uses a central, highly available business warehouse system that had a size of 19 TB in already compressed form before the start of the project.

"The BW system is very complex. All reporting for the approximately 5000 users with hundreds of queries per day runs through this one system. In addition, it accesses very many data sources."

explains Steffen Herold, SAP Solution Architect at HP, who is in charge of BW 7.31.

"Our customer takes full advantage of all the possibilities the system offers. For example, Navigational Attributes are used extensively to dynamically filter query results."

Since upgrading to version 7.31 in 2013, average query performance slowed from four to six seconds. The queries had grown over the years and were constantly expanded.

In addition, growing data volumes and system usage contributed to slower query performance. The introduction of a nearline storage (NLS) solution was intended to remedy the situation.

The choice fell on the software and consulting company DataVard and the OutBoard application. The SAP-certified archiving solution separates data according to operational relevance and stores it in a highly compressed (up to 95 percent) data store.

Decisive for the decision was, on the one hand, the integrated NLS Writer, which makes it possible to write to already archived data, and, on the other hand, the possibility to transfer Navigational Attributes into the NLS archive.

"An important argument was also DataVard's agility in being able to react to necessary changes and expansions accordingly at short notice"

explains Steffen Herold.

The NLS project presented a number of challenges: the implementation timeframe was tight at five months, and in addition, archiving had to be carried out during quarterly and year-end reporting without interfering with them.

Since the introduction of the NLS software required a change in the data model, extensive and complex test cases with many iterations had to be planned.

In previous changes, the most important reports had been tested manually using BEx export-to-Excel functionality (SAP Business Explorer). "A very time-consuming and error-prone procedure.

As a rule, the tester selects simple selections with one or two company codes instead of hundreds, and deviations are often overlooked," Steffen Herold explains the problem.

Due to the tight timeframe and the complexity of the BW system, the project managers decided to use the newly developed test suite DataVard Kate as an implementation accelerator.

The solution provides full support for usage data analysis, automation of test case and test data generation, and overall test management.

"The expectations were high. On the one hand, the solution should enable the early detection and elimination of software problems in the Q/A system, and on the other hand, it should ensure reporting data consistency and query performance."

says Steffen Herold.

Test execution was based on the three pillars of regression testing and data validation, performance testing, and testing of the data transfer processes (DTPs).

"Since the implementation took place during the year-end closing process, it was absolutely necessary to ensure that the existing logic and data were not changed"

explains Steffen Herold.

It was also important to test query performance in three variants.

"The measurability of the performance was a great help for us. We could see at a glance when times were off and adjust parameters or optimize queries."

Steffen Herold explains.

Since changes were made in the look-up logic (Abap code), the DTPs had to be tested. The SAP specialist found the repeatability of the test cases and the support for troubleshooting to be a great relief.

"Once the variants are determined, the tests can be restarted within seconds."

With the help of selections, a problem can be narrowed down further and thus the implementation can be checked step by step. In this way, it was also ensured in the productive system that everything behaved as in the tests in the Q/A system.

Another highlight: Test case generation can also be automated. The solution analyzes which queries are most frequently used with which selections, creates a HeatMap and generates the test cases from this.

"The HeatMap is a great way to map reality by using exactly the test cases that are really used"

Steffen Herold describes the advantages.

Test automation significantly increased test volume and depth.

"Using Kate, we were able to do extensive testing and test 110 queries in variants, for a total of 320 test cases. With manual testing, only the top 25 queries would have been possible."

estimates Herold.

In addition, the automated runs made it possible to ensure 100 percent data integrity.

"In a resultset of 100,000 numeric values, if one value is not correct, it is easily lost in the manual test"

explains the project manager. Those involved were also enthusiastic about the time savings:

"Without Kate, the implementation would not have been possible in this timeframe, bugs might not have been discovered and carried over into the production system"

says Steffen Herold. The NLS project with OutBoard also met all its goals: Overall, the BW system was reduced in compressed form from 19 TB to 11 TB.

"With the help of OutBoard, 2.5 TB of data could be archived. They are now compressed to 600 GB in the NLS archive".

sums up the SAP specialist.

Average query performance has also improved significantly following the introduction of NLS and as a result of other activities such as housekeeping, query and BW optimization, and BWA tuning. It is currently stable at three seconds.

"Like us, our customer is very satisfied with the results from the overall project. The NLS solution was implemented on schedule without any disruptions, the system size was significantly reduced and query performance was stabilized."

summarizes Herold.

Everyone involved was particularly enthusiastic about the possibilities offered by Kate.

"We estimate that the testing effort in the project has been reduced by at least 160 hours. In this sample calculation, we have calculated with only one run of 320 test cases and a manual effort of 30 minutes per test case."

says Herold.

In order to be able to use the effort and cost savings from test automation beyond the project, the software may soon be used permanently in the company.

At a glance

Challenge: growing data volumes, decreasing query performance

Solution:

• 2.5 TB of data archived, compressed to 600 GB
• Query performance improved from 6 to 3 sec.
• BW system reduced from 19 TB to 11 TB (compressed size)

System:

SAP NetWeaver BW 7.31, 5,000 users

Characteristics:

Highly available, central system, complex because of many data sources, many users. Customer exploits all possibilities offered by BW systems (e.g. navigational attributes).