Auto-migration with explicit trigger rules and safe cutover

Built for engineers who want deterministic behavior, and finance teams who need transparent, auditable cost controls.

How Auto-Migration Works

Step 1

Pulserun scans provider pricing and availability continuously.

Step 2

Migration triggers only when effective savings exceed policy threshold (default 15%), accounting for move overhead.

Step 3

State is checkpointed and target capacity is validated before traffic moves.

Step 4

Workload is transferred to better-priced matching capacity with orchestration and health checks.

Step 5

Job resumes from checkpoint; policy continues monitoring for future opportunities.

If a 24-hour H100 run starts at $3.00/hr and migrates to $2.10/hr after hour 4, expected savings are about $18 for that single run.

Checkpoint-first migration means no full-job restart. Teams keep progress and reduce failed overnight runs caused by manual re-launches.

Next: combine migration with preemption handling.

Create a free account and launch your first GPU instance in minutes.