A quick reminder on the workshop.

Why should you attend?

Most developers understand how to use a distributed database, but very few understand the ‘black box’ of how they maintain consistency under the hood. This workshop is designed to bridge that gap by building those internals from scratch. Following are some key points we will be focusing on.

Production-Grade Storage: We move beyond simple in-memory examples. You will implement a versioned storage engine using RocksDB, focusing on how to handle Timestamp-based versioning and write-intents.

Mastering Time without Atomic Clocks: You’ll gain a deep, intuitive understanding of Hybrid Logical Clocks (HLC)—the same technology powering modern giants like YugabyteDB and CockroachDB. You will learn how to achieve external consistency even when physical clocks drift.

Snapshot Isolation without a Central Oracle: Snapshot isolation is generally discussed with a central timestamp source (a Timestamp Oracle). However, databases like YugabyteDB don’t depend on a central source; they use causal tracking provided by Hybrid Timestamps. We will implement the key properties of Snapshot Isolation without depending on any central service.

Deterministic Debugging with TickLoom: We use TickLoom, a framework I built for Deterministic Simulation Testing (DST). It allows us to pause time and reorder network packets to catch ‘impossible’ distributed bugs that traditional testing would miss.

It was fantastic to host the offline workshop in Pune recently. Seeing a room full of senior engineers diving deep into the nuances of Timestamp-based versioning with RocksDB Storage was a highlight. Looking forward to more such engaging discussions with a wider group.

Building a distributed transactional database

{{ gettext('Draft') }}