Announcing Priority & Fairness for Temporal
summary
Temporal ships Task Queue Priority & Fairness that reorders async matches so critical work runs first and prevents noisy tenants from starving others. Priority uses separate in‑memory queue readers per level (default five priority levels) and merges heads; fairness implements stride scheduling with pass numbers and scalable counting via a count‑min sketch to handle millions of tenants. Demos show high‑priority tasks finish first, fairness evens dispatch rates, small tenants see big latency wins while heavy tenants take a modest penalty.
AI-written summary · about
chapters (7)
- 0:00Introduction to Priority and Fairness
- 1:28Problem Statement and Use Cases
- 7:03How Matching Service Works
- 13:06Implementing Priority
- 17:02Implementing Fairness
- 34:00Demos of Priority and Fairness
- 41:52Future Work and Q&A
related talks
18mScaling a Distributed Priority
24mMulti-Tenant messaging systems
40mEverything Will Flow
35mSIG Scheduling Deep Dive
50mThe end of event-based architectures? Orchestration with Temporal
26mTemporal: A Better Way to Build Modern Applications
23mDesigning a Workflow Engine From First Principles
46mBackground Jobs — What to Implement Them With and Why We Chose Temporal