Spot Instances

Caltech's Computational Biology group saved 70%, with 2x faster results.

Their team of scientists use Cedana to automate spot instances. No code modifications, configuration changes, or babysitting workloads.

Unbreakable, stateful reliability

Cedana makes sure you never lose work. Long-running, stateful workloads are automatically resumed on a new instance through revocations or failures. Your workload doesn't lose progress, and you don't waste time babysitting jobs.

Skyrocketing hosting costs?
If you're constantly adding servers to keep your website running and annoyed at your bill, Cedana automates spot instances to save up to 80% with zero code or config changes.
‍
Long-Running Batch Jobs Stuck on Expensive On-Demand?
Migrate lengthy batch workloads to spot instances. Cedana ensures they finish reliably, even through failures/revocations.
‍
Wasting Time Tuning Karpenter or Cluster Autoscaler?
Forget tweaking scaling parameters, setting complex PDBs, or configuring SQS queues. Cedana eliminates the guesswork of over or under-provisioning and the operational overhead of traditional spot management.
‍
Constantly Babysitting & Restarting Workloads?
‍Cedana transparently checkpoints, migrates, and resumes your jobs. No more manual checkpointing or restarting failed tasks. We provision exactly the nodes needed, when they're needed.

Automated

Live migrate GPU workloads before failures happen while system-level checkpoint/restore capabilities ensure no lost-work even during mid-epoch failures - even on large multi-node clusters.