Sharding, serverless, and scaling down

Designing Data-intensive Applications with Martin Kleppmann

Topic 7 of 14

Published Apr 22, 2026 · Clip 38:59–42:48 (3:49)

Martin explains that very high scale is still hard because sharding leaks into application code even with cloud services, but the cloud has greatly helped the low end: serverless enables cheap, lightweight, low-load services. On sharding, he argues it's more hardware power than cloud that lets more workloads run on a single machine; sharding across machines is less pressing though not gone, and replication remains relevant for fault tolerance.