Clustering in Orleans
Reuben Bond explains how clustering works inside Microsoft Orleans. Covers how application instances discover each other through a shared membership table (Redis, Azure Table Storage, Cosmos DB), health monitoring via ping/pong probes arranged on a consistent hash ring, the voting mechanism that declares unresponsive nodes as down, handling the distinction between crashed and slow nodes, heterogeneous clusters with different grain types, and indirect probing strategies inspired by the Lifeguard research paper.
Watch now →