What software and hardware do you need?
Based on the business requirements, the user groups and the architecture, we determine what tools we need in order to collect the data and to create reports and analysis.
Distribution of data to users
We also determine what tools we require for the distribution of data and reports. In addition to the specific Business Intelligence tools mentioned in another post, we will also need a ‘relational database management system’ (RDBMS) for the storage and retrieval of the data.
A long and short list
During this stage, we develop the selection criteria for the tools we require and we draw up a long list and a short list (with up to four suppliers) so that we will only invite relevant suppliers to display their products. The selection criteria are subdivided into ‘need to have’ and ‘nice to have’ to reduce the chances that good products will be rejected just because they score poorly on aspects that turn out to be relatively unimportant. It is also wise to find out what functionality is – or is not – essential for users.
Minimal coverage
The selection criteria should at least cover:
- the desired functionality
- ease of use for both developers and end users (with a specific focus on managers)
- maintainability, scalability and expandability
- testability
- tool compatibility with existing environment
- rough price estimate (per CPU or per source system type)
- compatibility with (support of) standards such as ANSI SQL, MDX and XML
- support for the existing infrastructure, such as operating systems, networks, platforms and databases
Start a proof of concept
We can start a so-called ‘proof of concept’ (POC) – a pilot project – and ask two or three suppliers, who meet the essential criteria, to show their solutions. During such a POC, it is wise to test various scenarios and to assess the tool on aspects such as:
- the processing of large volumes of data
- whether large numbers of users can perform actions simultaneously
- large number of ‘members’ in dimensions
- its connections to the source system and the exchange of meta data
- the degree of openness of the meta data repository
- the handling of historical data
- whether slowly changing dimensions are supported
- the possibility to use predefined, complex indicators
- the financial stability of the supplier
- the exact price and possible room for negotiation
- how the tool copes with status information such as inventories, subscriptions, prices and other data that are not additive over time (so-called semi-additive indicators)
Tools shouldn’t effect the architecture
It is highly desirable that the selected tools do not affect the architecture; however this is not always possible. Sometimes adjusting the architecture is unavoidable. This may be the case if the organization (internationally) opted for a standard solution in the area of Business Intelligence and this tool is not completely compatible with the designed data warehouse architecture.
Speed of the discs is important
During this phase, we also select the hardware on which the tools will run. Therefore, we will need to determine what capacity we require. An important aspect is the processing speed of the discs on which the data will be physically stored. This is often a major bottleneck when it comes to loading the data warehouse, reporting, and analysis. These processes often involve large amounts of data.
Comment on this post by Daan van Beek