Bridging the Chasm between Science and Reality
Abstract
When a research-prototype of a database management system becomes a product in the market, then more insight is required from the actual workloads to steer its industrial hardening. Unfortunately, few customers are willing to share their database schema, data samples, query load and execution traces for business and legal reasons. The technical challenge then becomes, how to enable customers to share their workloads in such a way that it will not leak sensitive information, while still provide sufficient information to allow DBMS researchers and developers to assess and improve their technology. In this paper, we report on ongoing research to address this challenge in the context of MonetDB1, as it is increasingly adopted in the enterprise market. This paper sheds light on the importance of a good profiling tool during system construction and the techniques deployed to gain permission from customers’ legal departments to share profiling traces captured on live production systems.