Introducing preemptible compute: the same compute, half the price

Summary

Together GPU Clusters is introducing preemptible compute in public preview for Kubernetes clusters in all regions, offering interruptible GPU capacity at a flat 50% of the on-demand rate. Preemptible nodes use the same NVIDIA accelerated compute, can be reclaimed when unused capacity is needed, and receive a maximum five-minute drain window for checkpointing and graceful exit. Clusters retain the requested preemptible target and automatically refill as capacity returns, while usage is metered every one to two minutes and billed sub-hourly. The feature suits resumable, retryable, or requeueable work such as experiments, batch inference, and temporary bursts, but critical services, uncheckpointed multi-day runs, and strict-SLO serving should remain on standard nodes; users must schedule eligible pods with the together.ai/compute-class=preemptible label.

Takeaways

1. 1

   Preemptible capacity is billed at a fixed 50% of the on-demand rate, with usage metered every one to two minutes and charged sub-hourly.

2. 2

   A reclaimed node has a maximum five-minute drain window; checkpointing can be triggered by SIGTERM, while Kubernetes events and API lifecycle events expose the preemption notice.

3. 3

   Preemptible nodes are intended for resumable, retryable, or requeueable workloads. Coordinators, login pods, serving replicas, and other critical components should stay on standard nodes.

FAQ