Post-Relational Databases Flying Under the Radar

In a SQL world, post-relational databases are holding their own.

NEWPORT BEACH, Calif.-In the modern IT world, SQL and XML have become such key components in the database sector that it's difficult to remember a time when they weren't part of the architectural story.

However, in the late 1960s and early '70s, database architectures with flexible field sizes, nested tables and loose-data typing options were winning market share and mind share. While these earlier systems lacked the tagging that in many ways defines XML as XML, they behaved in a remarkably similar manner in all other respects.

Because these systems-the current terms to describe them are "post-relational" or "multivalue"-don't get a lot of press coverage, it is easy to write them off as one of the many noble experiments that have fallen by the wayside. The IT industry is littered with better mousetraps that eventually disappear.

However, post-relational technologies haven't fallen. There are many companies that still offer such database products, from such tech-sector mainstays as IBM to lesser-known companies such as InterSystems, Northgate and jBase. At the International Spectrum Conference here in late March, seven of the major database vendors in this arena presented their wares, touted new partnerships and detailed plans for expansion.

A quick look through the conference agenda reveals the same topics that can be seen at a relational database conference: how to develop robust Web-to-data integration, change control management, security, document management and all the other usual subjects.

In addition, just like their better-known counterparts in the SQL world, these databases have strongly partisan supporters.

In reviewing the information at the event and talking with participants at the conference, here's the summary of the partisan case for why a business might consider post-relational databases in addition to-or instead of-RDBMS (relational database management system) technology:


SQL scales exceptionally well when scale means increasing the number of users without loss of speed. Post-relational systems scale exceptionally well when scale means increasing the complexity of the application.

The secret is in the data structure. Due to the XML-style nesting, data integrity is inherent in the model. This requires fewer computational resources to check and protect the completeness of the data. Additionally, nesting allows for dramatically fewer reads to achieve the equivalent volume of data retrieval. That means fewer read cycles and therefore longer MTBF (mean time between failures) relative to the volume of business. It also means that a programmer or analyst can see the primary relationships by looking directly at the data, without actually needing to see the schema.

IBM has documented excellent throughput with thousands of active users, so even user scaling works well in these environments.