OpenStack Swift is designed to make the most durable and available object storage system possible from your commodity hardware. From a few terabytes to dozens of petabytes and beyond Swift uses a consistent hashing ring to ensure data storage locations are dispersed to ensure failure-resistant operations of your clusters running at scale.
Swift operators are ring masters. They interact with them constantly to maintain order in their clusters, adding new capacity to power their growing cloud, removing or replacing old and busted disks and nodes, or creating new powerful differentiated storage offerings based on their clusters unique characteristics, geography or features.
In this talk we want to give you a deep dive into the rings: low level details how initial placement and rebalance works, best practices when designing cluster topology and ring management, and what you need to know about recent and upcoming changes like overloading, increasing partition power and composite rings.
Internals of the Swift rings and how they affect cluster layouts.