Cache Database

Last night at the CJUG the presention was by InterSystems on their Cache database (accent on the e but I’m not playing with high ASCII here.) I’d never heard of it, but apparently there are certain industries where it gets used a lot and the company brings in $150M every year because of it. In a nutshell, the database is not stored in typical DB tabular form but as a series of untyped multidimensional arrays. It exposes both a SQL interface and an object oriented interface, but both of those involve translation layers. This makes some things, like joins, close to a constant time operation. They made some pretty bold claims about DB speed, reliability/uptime and time to deployment. The sales engineer did a demo of their IDE/data modeler application and it seems quite cool. You model your schema with something a lot like Rational Rose or TOAD. It will generate the DB for you, and can very easily generate Java code to access that, C++, .Net etc. They also provide JDBC drivers so that if you prefer, you could treat it like a standard DB and go through the SQL interface.

One of my weaknesses is that I always fall for the pitches, and almost everything I hear presented sounds pretty good. This one is different, though. I think this really might be worth checking out further. They seem to have taken a path not unlike what I heard database curmudgeon Fabian Pascal propose, breaking the model of filling out tables on a disk in favor of actual low level storage that mirrors more closely the way data is accessed.