Running email infrastructure at scale involves solving practical distributed systems problems: storage, coordination, consistency and fault tolerance. This talk describes how to build and operate a large Stalwart deployment, using a 1024-node cluster as a concrete example of how the system behaves under scale. The number itself is chosen to illustrate scaling characteristics, but the methods discussed apply to clusters of any size.

The session will cover the technical considerations behind choosing a suitable storage backend, sizing a cluster based on expected load and redundancy requirements and evaluating different topologies for performance and availability. It will also dive into how cluster coordination is handled within Stalwart, including node discovery, membership management and state synchronization. The talk concludes with an introduction to orchestration and autoscaling, showing how to manage and adapt a Stalwart cluster in dynamic environments.

Attendees will gain a clear, practical understanding of what’s required to run Stalwart effectively in distributed production environments and how to reason about scaling it beyond a single node.