Strange Loop

Light and adaptive indexing for immutable databases

In an distributed setting with a large amount of append-only immutable data, the strategies needed to effectively index data change.

Recent research in database indexing is exploring so called "learned indexes", which uses machine learning to build predictive models of the data, and also adaptive indexes - where the indexes start out as small lightweight metadata that helps navigate the raw data, and become more fine-grained as the data is queried.

In this talk we'll discuss the engineering constraints we run up against in this setting, and how to solve them in ways that leverage the append-only and immutable nature of the underlying data.

Håkan Råberg

JUXT, Head of Research

Håkan Råberg is the Head of Research at JUXT. Since 2017 he has been part of the team working on JUXT's database product.