Cloud calibration
SWAT+ calibration and validation on AWS EC2
Dedicated EC2 per job — parallel MMPSO, live dashboard progress, auto-terminate when results sync.
Preview — Pro users by request while self-serve rolls out. · Access & allocations
FAQ
What AWS instance does SWATGenX use for calibration?
A spot c7i.8xlarge (32 vCPU / 64 GiB, Sapphire Rapids) by default. Larger swarms or big basins can use c7i.16xlarge (64 vCPU) or c7i.48xlarge (192 vCPU). Each calibration job gets its own dedicated instance — not a shared queue.
Does SWATGenX use spot or on-demand instances?
Both — chosen automatically from the estimated run length. Jobs that finish in under about an hour run on low-cost interruptible spot. Longer jobs run on a dedicated on-demand instance that AWS cannot reclaim mid-run, so a multi-hour calibration completes through validation instead of being interrupted partway. If a short spot job is reclaimed mid-run, SWATGenX automatically relaunches it on a non-reclaimable on-demand instance, warm-started from the last saved iteration, so the run still finishes. Either way the instance auto-terminates the moment results are fetched, so there is no idle compute.
How long does a calibration take on AWS?
Roughly 15-20 minutes for a small basin, a few hours for a mid-size model, and up to ~1 day for the largest continental basins on a single 32-core instance. Wall time scales with HRU and channel count, the calibration window, and the swarm size; bigger swarms or instances finish large basins faster.
Can I watch the calibration live and stop it myself?
Yes. Progress streams to your dashboard every ~20 seconds — objective metrics, the initial-stage hydrograph, and the global-best and per-particle convergence curves — so you can see how the swarm is converging in real time. You can stop the run at any point based on your own judgment of the fit (for example, once the global best has plateaued); SWATGenX keeps the best parameters found so far and shuts the instance down.
Do I have to manage any AWS infrastructure?
No. SWATGenX bundles your model with the pinned SWAT+ engine, launches the instance, runs the PSO swarm, streams live progress to your dashboard, fetches the results, and terminates the machine automatically — you only see the calibrated parameters, metrics and hydrographs.
Is cloud calibration available now?
It is in preview and offered to Pro users by request. The infrastructure is live and validated; contact us to enable it for your account while we roll out self-serve access.
Next steps
Last updated 2026-08-08.