An application-specific database

Abstract

Signaturesare evolving profiles of entities extracted from streams of transactional data. For a stream of credit card transactions, for example, an entity might be a credit card number and a signature the average purchase amount. Signatures provide a high-level view of data in a transactional data warehouse and help data analysts focus their attention on interesting subsets of the data in such warehouses. Traditional databases are not designed for such applications. They impose overhead for services not necessary in such applications, such as indexing, declarative querying, and transaction support. Hancock is a Cbased domain-specific programming language with an embedded domainspecific database designed for computing signatures. In this paper, we describe Hancock’s database mechanism, evaluate its performance, and compare an application written in Hancock with an equivalent application written in Daytona [5], a very efficient relational database system.