Monolithic vs. Logical Architecture: Which for the Win
Reading Time: 3 minutes

As any data leader will tell you, it is impossible to derive any value from data unless it is accessible and reliable, regardless of where the data resides. Unfortunately, organizations are far from achieving this goal, because their data is spread across multiple silos. This goes into many topics: standard vs modern data architecture, data democratization, but they get in the way of the core topic: what’s better, monolithic or logical architecture?

Overcoming Data Management Woes

There are two basic types of data architecture with which companies are trying to solve the data-silo issue. The first is monolithic architecture, a centralized architecture wherein all relevant data for analytics is first copied to a single central system, then transformed and prepared. The classic example is a data warehouse, which aggregates data from different sources into a single, central, consistent data repository to support reporting needs. Organizations find this approach compelling because it promises to keep all the data in the same place, eliminating silos, which in turn helps achieve consistent semantic security and data governance.

But it has proven to be extremely difficult for larger organizations to store all their data in the same place. This is where logical architecture emerges.

With this approach, data can be located in multiple distributed data processing engines, which may be specialized at different tasks. But on top of those disparate systems is a common logical layer, which in most cases is enabled by data virtualization, a modern data management technology that also provides unified semantic security and data governance. Logical architecture enables users to achieve consistency without needing to first consolidate all of the data in a single system.

A Monolithic or Logical Approach? 

The main difference between the two types of architecture is that with monolithic architecture, to create new data products, all of the required data needs to be copied into the central system and then physically transformed and combined with the data in different ways. This is neither an agile nor a quick process. But with logical architecture, users can reuse existing systems for a set of data products. And even if they decide to ingest the data into a new system for certain data products, when data needs to be exposed in different forms, it can be done without physically replicating data every time.

A second difference is that with monolithic architecture, users need to solve all of their analytics needs with the same system, the  central, monolithic system that supposedly contains all the data. But with logical architecture, users can have different processing engines optimized for different types of analytics. For example, they could use one system for data science and another for traditional business intelligence. Logical architecture provides these abilities while still enabling unified data operations comprised of security and data governance capabilities.

A third difference is that with logical architecture, consumers access data through a semantic layer that expresses the data in everyday business language. This is useful not only because it simplifies data access, but also because data can then be easily moved to a different system or even to the cloud, and it would still be findable and understandable, due to the semantic layer. These are some of the major benefits that logical architecture brings to modern data architecture.

In general, logical architecture is the stronger approach. Now let’s look at a specific use case: cloud adoption.  

Cloud Adoption: Logical Architecture for the Win

For cloud adoption, logical architecture again emerges as the better choice, and for several reasons. First, during the cloud migration phase, when systems are being moved to the cloud, logical architecture can help move data and workloads to the cloud, minimizing the impact on data consumers. This is due to the semantic layer mentioned above. Data consumers can continue to access their data using the abstraction provided by the semantic layer, so they don’t need to know where the data physically resides.

In multi-cloud configurations, data systems are distributed across different cloud providers and need to be integrated among them. In that scenario, the capabilities of logical architecture for enabling distributed data integration while still maintaining unified semantic security and data governance across all the clouds is very useful. And most importantly for organizations, logical architecture also decreases the overall cost of the cloud migration. Taking all these factors into consideration, in the context of cloud adoption, logical architecture offers significant benefits to organizations.

For more insights on this topic, check out our conversation with Alberto Pan, Executive Vice President & Chief Technical Officer at Denodo, on All Things Data!

Neha Gurudatt