Historically, RDBMS could manage the majority of data tasks. However, with the advent of management information systems, the need for Online Analytical Processing (OLAP) and column store databases became evident. Key value pair and document stores soon followed, addressing unstructured and semi-structured information needs. Finally the rise of graph databases has presented new opportunities and challenges for data handling, providing both OLTP and analytical functionality.

How can customers integrate all these databases capabilities into their IT landscape?

Stay with RDBMS

One approach is sticking with an RDBMS vendor that offers extensions for additional features necessary for modern data management. For instance, PostgreSQL has developed functionalities to handle documents through various plugins, including column and graph store features. However, these enhancements are often retrofitted into a system initially designed for relational data. Consequently, there may be inefficiencies at query time due to necessary data conversion processes. Additionally, because these functionalities are added on, they might not be as seamless or efficient as bespoke solutions.

Best-of-Breed Approach

Alternatively, the best-of-breed strategy involves selecting specialised databases, each tailored for a specific type of data. This approach allows the use of purpose-built technologies, such as MongoDB for document stores, Vertica for columnar storage, and Neo4j for graph data. While this strategy offers optimal performance for each data type, it also introduces complexity. Managing multiple database systems can be challenging and often requires sophisticated integration between the databases.

Multimodal Databases

The third option is adopting multimodal databases, designed from the outset to handle various data types within a single system. Microsoft’s Cosmos DB exemplifies this approach, aiming to support multiple models such as document, graph, and columnar data. Although promising in its versatility, multimodal technology remains relatively new and can present limitations in performance and maturity compared to specialised systems.

Choosing the Right Strategy

Selecting the appropriate database architecture is contingent on several factors, such as the specific needs of the organisation, the complexity of data requirements, budget constraints, and the technological expertise of the team. Companies must weigh the trade-offs between integration complexity, performance optimisation, and future-proofing their data infrastructure.

No single strategy suits all. Organisations must carefully evaluate their current and projected data landscapes to determine the most effective approach, ensuring flexibility and scalability in their data architecture.