Workspace Backups & Migration

A pod is an Environment (the base image, the packages you installed, the startup script) plus a Volume (/workspace: your code, checkpoints, and datasets). Stopping a pod captures both, confirms they succeeded, then releases the machine. There is no manual capture button and no schedule to set up: it happens on every Stop, whether you triggered it or auto-stop did.

The Environment side stays still between stops: your working environment is the pod's own business, and the only way to freeze a numbered version of it is to Share or Keep it. The Volume side keeps moving: while the pod is running, Aquanode saves your Volume in the background every minute or so whenever it has changed, so a Stop is never the only thing standing between you and lost work.

Captures are incremental: the first one carries the directory, later ones carry what changed. A capture taken seconds after another with no writes in between costs almost nothing, in both time and stored bytes.

This is the part that matters. Neither capture is welded to the instance that produced it. When you start a pod, its environment and volume restore onto whatever box you point it at: a different GPU model, a different region, a different provider entirely.

So "migrating" a workload is not a special feature with its own machinery. It is the same round trip you already use every day: stop the pod you are on, start it on the machine you want, and both captures restore onto it. Moving to another GPU while a pod is running does exactly this in one action, and never releases the old machine until the new one is confirmed.

Comparing prices across providers only helps if you can act on what you find. Without portable state, a cheaper GPU somewhere else is a rebuild: reinstall CUDA, re-pull weights, re-create the environment, lose a morning. That friction is what keeps people on whichever cloud they set up first.

A portable environment and volume are what turn a price difference into a decision you can make on a Tuesday afternoon.

Aquanode's own storage is the zero-configuration default and needs no credentials from you. Captures and restores work the moment you launch a pod. Capturing your data costs nothing extra beyond what it holds: storing it in Aquanode Storage is billed at $0.02 per GiB per 30-day month, metered per second. Writing to a bucket you own instead (your account, your region, your keys) carries no Aquanode storage charge and is part of the enterprise plan; book a demo and we set it up with you.

Storage fees are listed in the console separately from GPU cost, so a restore never shows up as a mystery line on a GPU bill.