• No results found

It is often difficult to find the most qualified BW consultants and internal resources to staff a BW project. SAP does have a BW certification process.

This is a test that is administered by SAP to determine some level of profi-ciency in the product.

In my experience, it is not helpful to use the certification process to deter-mine the level of the consulting resource during the interview process. I have interviewed and worked with various consultants who have been certi-fied (as am I).

I did not feel that the certification proves whether a consultant is any more qualified to implement BW. In some cases, the certified consultant resources were knowledgeable in the terminology but not in the processes and imple-mentation strategies. Therefore, do not assume that a certified consultant is any more qualified than one that is not certified.

The lack of knowledge in the process and implementation during interviews is likely the result of the many test questions and books available to help

Project Planning in BW

4

people pass the test and become BW certified. Thus, I do not recommend that you look at these criteria alone when determining whether a consultant is qualified to implement BW. It does help to highlight those consultants with a basic understanding of the BW product, but not necessarily those who would be a great resource on your project.

I am most successful in finding the best resources when I conduct a detailed interview of the candidate. This allows me to understand not only their product knowledge but also their communication skills. Because both are equally important, I ask a lot of very open-ended questions to try to get the interviewee talking about his or her specific experiences in one area of BW implementation.

When I first started working in BW, the resumes seemed to match with the skill sets of the interviewees very well. However, in the last several years, I have seen many resumes that are embellished beyond what I would ever have imagined. A resume comes across my desk for someone that I worked with in a previous project. I knew their role on the BW team, and I’m frus-trated when the resume lists many things that were not part of that role.

This trend seems to be a result of the many search tools that recruiters use to find consultants. Thus, consultants who were not closely working with CRM feel they need to list it on their resume to make sure that their resume comes up when CRM is searched in a job website. This is unfortunate, because I find myself asking consultants about things on their resumes that they can-not talk about with any degree of certainty.

This has made the interview process even more important. Of course, lean-ing on the other BW consultants I know has helped a great deal too. Because the world of BW is not that large, many times I can find someone who has worked with a particular resource and can give me some details about their background before I ever schedule an interview. This networking has allowed me to informally weed out those candidates who are less than qual-ified.

Interviewing BW candidates is not easy, especially if you are new to the product. I have included some interview questions that I like to use when I evaluate a BW consultant:

1.What is the difference between an DSO (also called DataStore Objects) and InfoCubes and why is this difference important?

This question is intended to make sure the candidate understands the basic architecture of BW. This is a very basic question with quite serious

rami-BW Interview Process 4.7

fications if the candidate does not understand the use of each of the struc-tures. The answer should clearly indicate that the InfoCube is the primary reporting repository for BW data, although the DSO can be used for reporting in some circumstances. The database structure of the InfoCube and DSO are completely different.

The DSO is a relational database structure with a user determined key and can have additive or overwrite capabilities. The InfoCube is a star schema design. The key is made up of a combination of the various characteristics in the dimension tables that establish dimension IDs used in the two cen-tral fact tables for reporting. The InfoCubes is only additive, there is no overwrite capability in the InfoCube.

2.What kind of things can be used to determine query performance bottle-necks?

Performance management is one of the more difficult tasks in BW. This is because if you understand what performance issues can occur, you clearly understand the underlying design of the BW system. I have found that the more a candidate understands performance in BW, the more he or she seems to understand the product and its interworkings, and thus the stronger applicant he or she is. Thus, I like to ask many different perform-ance related questions to determine if the interviewee understands per-formance management in BW.

This also makes sure if candidates understand it, they will incorporate this into their data structures in BW, which is clearly an important overall goal of any implementation. In this answer, the candidate should mention something about BW statistics and an ongoing process to monitor these statistics. There are also some various other performance tools, such as transaction RSRV. But for the most part, determining performance bottle-necks should come from the statistics InfoCubes.

3.What are some things that can be used to fix query bottlenecks?

This question is trying to get to the root of the performance questions to see how the candidates would actually apply their knowledge to a per-formance issue. Because there are many reasons for perper-formance issues, I do not let a candidate give just one answer. I look for several of the follow-ing answers:

Adjust the query, adding filter values to reduce the data retrieved

Check the query to see if the filter logic is using exclusive logic

Check the data model. If the query is going after an InfoProvider with massive amounts of datam reduce the volume by segregating the

Info-Project Planning in BW

4

Providers into smaller individual ones, and use a multi-provider or adjust the data model appropriately

Check to see if the query is running from an ODS/DSO. This is not the preferred InfoProvider for query performance; switch to an InfoCube.

This is because queries typically run faster from InfoCubes, and more performance enhancing tools are available on InfoCubes. If the query is running from an ODS/DSO, the candidate should mention adding indices to the ODS/DSO.

Check to see if the query is coming from a virtual InfoProvider

Implement BW Aggregates

Turn on/adjust OLAP Cache settings

Implement the BI Accelerator (available only in NW 2004s and above)

Implement partitioning

Implement compression

4.What is a navigational attribute and why would one use this in the BW design?

This question is posed to ensure that the interviewee knows about the con-cept of data being read from either the transactional data or the attributes of master data. A navigational attribute is a master data field that is an related field of master data that can be used for reporting. For example, transactional data is loaded into BW with the customer and a net sales value. The customer master data record has an attribute ”country“ that is populated on the customer master data, but not in the transactional data.

Although the transactional data does not have the country field, the data can be reported by country because the master data table for the customer does have the country.

This allows the BW report to show net sales by country without having this field in the transactional data. They need to fully understand that data read and accessed from the navigational attributes is stored in the master data and is dynamic. It changes with the master data itself. This provides an automatic restatement of the data each time the master data is reloaded.

In this question, I am trying to make sure that they understand the concept of using master data in their design. There are several ways of delivering reports in BW. If all the data comes from the InfoCube directly, this means that all reports are sourced in the transactional data.This can be fine if the users are only concerned about a snapshot of the transactional data at the time that the transaction occurred.

BW Interview Process 4.7

However, understanding that the user can also view the data as a reflection of the current master data attributes gives them a more dynamic approach to the data, allowing transactional data to be restated based on the current view of the master data attributes. A navigational attribute is a master data field that is an related field of master data that can be used for reporting.

For example, transactional data is loaded into BW with the customer and a net sales value.

The customer master data record has an attribute ”country“ that is popu-lated on the customer master data, but not in the transactional data.

Although the transactional data does not have the country field, the data can be reported by country because the master data table for the customer does have the country. This allows the BW report to show net sales by country without having this field in the transactional data.

5.What is the delta initialization process and why is it important? Is there a way to resend a failed delta? How?

This question is intended to make sure that the candidate has actually loaded data into BW on an ongoing basis. Sooner or later a delta will fail for various reasons including data issues, space issues, etc. The interviewee should understand that the initialization of the delta process establishes the delta queue in the source system. This delta queue is used to store and track the changing delta records.

If a delta fails, it is possible to resend this delta by simply backing out the failed delta on the BW system and using the delta-processing transactions in the source system, including Transaction RSA7, to resend a failed delta.

6.What is the change run? (It is also called the apply hierarchy change) Understanding the change-run job shows that the interviewee has done some work managing the batch processes of BW. Master data comes into BW in an inactive state. Changes to master data are not immediately reflected in the master data values. This job simply takes the new master data values and moves them to the active status. This allows the changes in master data to be reflected.

The job also has one other important role. It adjusts any BW aggregates that have navigational attributes to the new master data value. Simply put, it does for the aggregates the same thing that it does for the master data. I have found that consultants who understand both of these parts of the change run have usually worked hands-on with aggregates, batch, and master data. This is a good sign that they understand the most important aspects of the BW design.

Project Planning in BW

Related documents