A Handbook for Building an Approximate Query Engine.
Barzan Mozafari, Ning Niu · IEEE Data(base) Engineering Bulletin · 2015
There has been much research on various aspects of Approximate Query Processing (AQP), such as different sampling strategies, error estimation mechanisms, and various types of data synopses. However, many subtle challenges arise when building an actual AQP engine that can be deployed and used by real world applications. These subtleties are often ignored (or at least not elaborated) by the theoretical literature and academic prototypes alike. For the first time to the best of our knowledge, in this article, we focus on these subtle challenges that one must address when designing an AQP system. Our intention for this article is to serve as a handbook listing critical design choices that database practitioners must be aware of when building or using an AQP system, not to prescribe a specific solution to each challenge.