Lynn Wheeler can't seem to get away from databases. A retired IBMer, he's using a concept he worked on some 30 years ago to help governments and other institutions connect and merge information they already have.
In the age of cloud computing and nanotechnology, databases may seem pretty mundane, but they impact our lives daily. Whether you are grabbing cash from a machine or visiting a store that knows which popular items to put on the shelf, such activities will typically involve large databases.
ATMs and other contemporary applications rely on relational databases, and relational database structures are relatively static.
Getting real.
"A relational database only works with clearly defined relationships," said Lynn, explaining that various sets in the database must be predefined before any data is entered. That works fine for things like ATMs which involve well defined and understood elements - PIN, account number, name, etc.
"In the real world, relationships are less structured and not homogeneous," said Lynn.
Recently, he said, Maryland tried to use a federal database that predefined limited things known about people. The state had a group of peaceful demonstrators that it wanted to enter into the database, but there was no category for "peaceful demonstrators." The state was forced to choose the existing terrorist category, even though that wasn't accurate.
Needless to say, that wasn't good news for the reputations of peaceful demonstrators. Even worse, the information was shared with other agencies, making it hard to correct or eliminate.
Creating database for complex, real-world information can be an arduous undertaking. According to Lynn, it can take man-years (in one case involving very complex information, taking nine months elapsed time to create the relational definition).
His semantic approach
Lynn works with an approach called a semantic network database. It builds on a concept developed inside IBM in the 70s and 80s, he said. "More than a decade ago, I started a brand new implementation," he said, noting that the patents on the original work have all expired.
Rather than relying on predefined uniform definition, a semantic network database can dynamically add arbitrary relationships between any piece of information and any other piece of information. Even better, tables from traditional relational databases can be easily imported (and merged).
That's important because many entities already have lots of relational databases. For example, hospital records are stored in disparate databases, each having its own predefined, uniform structure. They don't share the same predefined form and thus can't talk to each other. Attempting to reconcile and/or merge information from different relational databases is frequently a lengthy and onerous undertaking.
More effective security
Similar valuable information lies in databases scattered among government agencies. Being able to access information about a person in databases run by the FBI, Immigration and the Border Patrol, for example, "would prevent a lot of things that could otherwise happen," said Lynn.
The issues faced by the intelligence community were included in a 2006 article
DataMining "Disrupts and Enables" (alternatively from wayback machine).
Just what he's offering is reflected in Lynn chose the name Dynasty for his company to reflect the richness that can be found in relationships.
The challenge facing Lynn and his firm is that most database users are "very much oriented to commercial off-shelf products." He's looking for customers who "attach significant value to what they can't do."
Finding those customers isn't easy - but perhaps they'll find him. After all, Lynn can't seem to get away from databases.
Greater IBM gone 404
https://web.archive.org/web/20110203034355/http://www.ibm.com/ibm/greateribm/
may forward to
https://www.ibm.com/gr-en/about