10M+ Tasks per Day
Hyrex is engineered to push sustained, large-scale throughput. With stateless workers, short leases, and lightweight coordination in Postgres, you can process 10M+ tasks/day reliably.
Scale horizontally by adding workers; Hyrex keeps distribution fair and fail-overs safe. Use batching and per-task concurrency limits to squeeze the most out of your fleet.
Throughput math
- 10M/day ≈ 116/sec: Roughly 116 tasks per second sustained.
- Workers × rate: 50 workers at ~2.5 tasks/sec each reaches ~125/sec.
- Latency vs batching: Small batches (5–20) amortize I/O without hurting P99.
Tune batch sizes and concurrency per worker based on task cost and downstream limits.
How Hyrex sustains volume
- Short, renewable leases: Fast detection of stale ownership for steady progress.
- Fair rebalancing: Workers periodically redistribute claims to avoid hotspots.
- Efficient polling: Lightweight selects keep queues hot without hammering the DB.
- Deterministic retries: Jittered backoff smooths spikes during partial outages.
Recommended patterns
- Stateless tasks: Keep tasks idempotent; store progress in your DB/obj store.
- Batch I/O: Coalesce external requests (e.g., 10 writes at a time) to reduce chatter.
- Protect dependencies: Use per-task concurrency and rate limits for fragile APIs.
- Backpressure: Let queues absorb bursts; scale workers to drain sustained backlog.
Operate with confidence
- Watch the backlog: Scale workers if pending tasks trend upward.
- Track dequeue latency: Rising time-to-start indicates undersupply.
- Use Studio dashboards: Inspect throughput, failures, and retries in real time.
Need more scale?
Hyrex Cloud can handle 10x–100x more throughput than the Postgres-based version, with the same developer experience and operational model.
When you outgrow a single Postgres-backed cluster, move to Hyrex Cloud to push even higher sustained volumes without changing your code.
Explore Hyrex Cloud