Architectures for analytical data platforms

It is essential to consider the various architecture models for analytical data platforms when developing a data strategy. It should be noted that the architecture models address different aspects to varying degrees.

While architectures such as classic data warehouse or data lakes and data lakehouses imply certain forms of data storage, approaches such as data fabric or data mesh are more open and require more specific details regarding types of data storage.

Data mesh architecture, in turn, exists in various forms that are discussed academically, but leaves the question of infrastructure and data storage open. Here, questions of decentralized organization and data responsibilities are brought to the fore. This must therefore be combined with other architecture models (e.g., data warehouse, data fabric, or data lakehouse) when defining a strategy.

We have relevant project experience with a wide range of architectural models, from classic relational models in data warehouses to delta files in data lakehouses, as well as with domain tailoring and the definition of cascading data products in data mesh model.

The appropriate combination of different architecture models and the components of these models is the content of the initial formation of a data strategy with organizational elements and technological decisions, right through to the concrete selection of analytical platforms and tools. In practice, the approaches are always combined in a meaningful and customer-specific way, e.g., a data lakehouse implementation with a data mesh approach to organization and governance. We are familiar with the critical variables in decentralized responsibility and centralized support functions or infrastructure and would be happy to work with you to develop the right concept. We call this best practice and know that these strategic decisions are far-reaching and relevant to the success of the project.

When choosing technologies and tools, we place great importance on cost optimization, scalability, and transportability, the latter of which reduces dependence on specific cloud providers and technology providers.

Even though models such as data lakes and data lakehouses are heavily influenced by cloud-based technologies and providers primarily rely on the cloud, at DATA MART we have the project experience to implement these architectures completely on-premises for customers if this is appropriate for cost reasons or due to very high regulatory requirements.

Get in touch with us.