11.3 Access Versus Storage

Rather than bringing the data to a centralized data warehouse, accessing it remotely is just as effective. So, rather than replacing and consolidating isolated data sources, as is the case with many government databases at the federal, state, and local levels, remotely referencing this shared data works as well and is more cost-effective. Sharing data can be difficult in a heterogeneous database environment. There are issues with shared versus replicated data, data consistency, and performance. But there are many applications, such as data mining, that require high flexibility and low maintenance costs, yet impose modest performance requirements for data. Sometimes there are good reasons to leave data in different database systems, whether historical, functional, or cost-related.

When a company or a government agency sets out to build a decision-support system that will make all its data accessible, it does not necessarily mean that a query about a customer or an individual will be restricted to a single database. In fact, the GRASP system allowed users to access multiple databases with a single query. These diverse data silos contained telephone records, vehicle registrations, driver's licenses, public record filings, change of address records, credit bureau data, real estate records, and, of course, internal government databases. No warehouse at that time could have stored the billions of records that the GRASP system accessed: that's because we opted for access rather than storage. We built a network rather than a data warehouse.