Fair & Square: Building Bulletproof Multi-Tenant Job Queues on Postgres + Kubernetes
Presented by:
Caleb Schoepp
Caleb Schoepp is a senior software engineer at Akamai and the co-champion of the WASI OTel proposal. Previously he interned at Microsoft, UnifyID and Resemble AI. Caleb enjoys hacking on cloud infrastructure and learning the ins and outs of WebAssembly. Outside of work he enjoys playing guitar, spending time with family, playing hockey, and racing in triathlons. Caleb holds a BSc in Computer Engineering from the University of Alberta. He lives in Edmonton, Alberta.
No video of the event yet, sorry!
When you run jobs on behalf of thousands of tenants it’s only a matter of time before one noisy neighbor floods the system, starves everyone else, and brings your queue to its knees. This talk is about building a bulletproof multi-tenant job queue using tools you already have.
We'll start with the database. Postgres is a perfectly good job queue if you know the tricks: SELECT ... FOR UPDATE SKIP LOCKED for safe concurrent dequeues, and how to avoid the lock contention and wasted work that naive implementations hit. Then we'll scale the workers out with Kubernetes and KEDA, autoscaling pods directly off queue depth.
The interesting part is the fairness mechanism. We'll bring an algorithm, Start-time Fair Queuing, out of the papers and into a working application showing how fairness is achieved. We'll also cover tenant isolation and the security model for running untrusted work. You’ll walk away with practical tips on how to adjust the job queue design for your own workloads.
- Date:
- Duration:
- 40 min
- Room:
- Conference:
- PG Down Under 2026
- Language:
- Track:
- Development
- Difficulty:
- Medium