Static data structures have been at the heart of data processing tools since the dawn of computing, but they have always limited the flexibility of the organization leveraging the data. Recently, the rise of flexible formats like JSON have led to schemaless data as an attempt to increase agility. However, schemaless data have proven difficult to work with, because of hidden rigid structure in the form of implied schemas.
EnterpriseWeb addresses the problems of both the inflexibility of structured data as well as the impracticality of schemaless data, by enabling schemaless data to be dynamically and logically structured.
From the fixed-length fields of the 1950s, to the relational structures of modern database management systems, to the semistructured data formats XML and JSON, the structure of our data has always informed code about how it should be processed. Data are defined by their relationships, and we used to hard-code those relationships into rigid structures. That approach allows only one static view, which is difficult to work with, and even more difficult to change. Nevertheless, such rigid data structures – and the models that represent them – are an integral part of enterprise information management.
Traditional relational database management systems (RDBMSs) exemplify this point with their static entity-relationship models (ERMs) and tightly interconnected data structures. XML improves this situation slightly, allowing semi-structured information, but schemas still constrain flexibility and performance. With both approaches, fixed definitions, views, and reports limit the ability for businesses to freely transform information into insight and become obstacles to systemwide change.
The Rise of Schemalessness
This challenge of inflexible data structures has given rise to schemaless data. With JSON in particular, we can create whatever data structure we like when we author data. We don’t have to shoehorn data into rigid data structures, thus allowing every record to have its own structure.
But there is a problem with schemaless data. Consider this simple task: how do you create a query for all the addresses in a particular Zip Code if every record has a different name or format for Zip Code? Schemalessness, after all, isn’t magic – even schemaless data require some kind of metadata so the code will know how to process such information, what software development guru Martin Fowler calls an implied schema.
Implied schemas represent the structure inherent in any data record. If each address record has its own format, then that format provides the implied schema for that record. Dealing with implied schemas thus falls to the developer, who must figure out how to code software to process these implied schemas, which are different for each and every record.
In Fowler’s tutorial on schemalessness, he explains the pros and cons of implied schemas. Despite acknowledging the power of schemalessness to support more flexible and responsive user experiences, he recommends avoiding it and implied schemas for developer convenience. Good advice with respect to traditional software, but the world of data is changing. Today we live in an increasingly schemaless world, where more often than not, the structure of our data is fluid or nonexistent.
Raising the Discussion to Dynamic Schemas
Fowler makes it clear that in the past it has been impractical from the developer’s perspective to work systematically with schemaless data, because implied schemas are difficult to deal with. After all, structure is itself useful, and isn’t the problem per se. Rather, how to avoid the limitations of static structure without falling into the trap of unmanageable schemaless data that is the real challenge.
EnterpriseWeb’s unique approach to modeling solves this critically important challenge by leveraging dynamic schemas that have flexible, metadata-driven relationships with underlying information. Using metadata this way separates concerns, letting people consider relationships from multiple perspectives, rather than from a single static point of view. In addition, it’s now possible to change and extend metadata to meet diverse business needs without disruption.
Instead of settling for complex ERMs with their inflexible, tightly coupled data structures or dealing with the coding complexities of implied schemas, developers can project dynamic schemas from the metadata simply by writing different transformations. As a result, dynamic schemas are developer friendly and dynamic – a welcome change from the difficult problem of schemalessness.
Add an Agent for Performance
So far so good, but how do we build software to process all such data in a general way, freeing ourselves from custom coding for implicit schemas? The solution is an intelligent agent.
EnterpriseWeb’s intelligent agent, SmartAlex™, is a distributable transaction manager that resolves dynamic schemas for each interaction. Every human or system client interaction is a request for SmartAlex to interpret dynamic schemas (as well as other models and additional metadata) and translate them to a context-specific set of resources in order to construct a custom response.
This Agent-Oriented approach maximizes performance for such dynamic computing. In the background, SmartAlex handles all run time connection and transformation details, sparing programmers from manually integrating resources for varied and unanticipated uses, greatly improving IT productivity while enabling business agility.
SmartAlex logs all system events, indexes all new and updated resources, and tags all changes in relationships for detailed and navigable audit history. This practice creates a feedback loop as SmartAlex leverages the same indexed logs to guide its execution. Data, code, and user interface components, as well as connectors for federated services, systems, databases, and devices, can be updated or replaced without breaking related apps and processes – as SmartAlex is ‘aware’ of the changes. In this way EnterpriseWeb supports real time exception and change management for resilient solutions that can evolve naturally.
The EnterpriseWeb Take
Schemalessness was a reaction to the limitations of structured data. People struggled with the constraints of static structure, and figured that if they simply got rid of structure, then the problem would go away. But this move was merely a shell game, as the limitations of fixed schemas shifted to implied schemas, now without the benefits of structure to inform the code responsible for their processing.
The solution is to raise the level of abstraction, and instead of arguing over fixed vs. implied schemas, to work at the dynamic schema level. Such an approach is model-driven, allowing application designers to build models that capture their data structures, and allowing an intelligent agent to use the metadata each model represents to meet the specific needs of each interaction. The real lesson here is that the solution to resolving the challenge of schemalessness combines both dynamic schemas and the action of the agent. Stay tuned to my next newsletter for more information.