Most of what teams spend on SageMaker Endpoints in non-production is spent while nobody is watching. In a development, staging, or QA account these resources are billed for every hour they exist, but the people who use them work a fraction of those hours. Nights, weekends, holidays, and the long tail of “we’ll get back to that environment next sprint” all meter at full price.
ZopNight closes that gap by scheduling your SageMaker Endpoints to run on your team’s hours and stop the rest of the time, without touching your data and without a migration. It is the most direct lever in FinOps, and for teams comparing tools it is where ZopNight pulls ahead of dashboard-first platforms like CloudHealth.
Why SageMaker Endpoints cost more than they should
Cloud providers bill SageMaker Endpoints by the hour whether or not anyone is using them, and most non-production fleets default to running 24/7 out of habit rather than need. At typical on-demand rates ($0.05 to $5.00 per instance-hour) an always-on instance quietly bills the same on a Tuesday afternoon as it does at 3 AM on a Sunday.
Here is the shape of it. Say a box runs at a typical on-demand rate of $0.17 an hour. Left on around the clock that is about $124 a month, because a month is roughly 730 hours. Confine it to a single-shift work week, about 50 hours, and you pay for 50 hours instead of 730. Your own rate and hours will differ, but the ratio is the point: in non-production, most of the meter runs while nobody is working.
The instinct is usually to reach for a smaller instance type. But rightsizing only helps a resource that is genuinely too big; it does nothing for a correctly-sized resource that simply runs when no one is around. The larger, easier win is refusing to pay for the hours nobody is working, and it carries none of the performance risk of down-sizing a box that might spike tomorrow.
Stopping and starting SageMaker Endpoints safely
ZopNight handles the off-hours schedule of the endpoint instance count on every SageMaker Endpoints in scope, in dependency order so nothing comes up before what it depends on. A scheduled stop preserves your data exactly as a normal power-off would; ZopNight never terminates or deletes the resource, and idle detection watches CPU, network, and disk signals to surface the SageMaker Endpoints that are running but doing nothing.
In practice the setup is: Connect your AWS account; ZopNight discovers SageMaker endpoints along with Invocations and ModelLatency metrics; Idle-endpoint rules read the Invocations metric to separate a truly idle endpoint from one with real but intermittent traffic; Schedule non-production endpoints off outside working hours and back before the team needs them; Layer in rightsizing recommendations for endpoints that are provisioned larger than their traffic needs.
If you run SageMaker Endpoints you probably also run SageMaker Notebooks, SageMaker Hyperpod, and scheduling them together is where the dependency ordering earns its keep. Related reading: scheduling and idle detection.
How ZopNight schedules SageMaker Endpoints
The loop that does this is deliberately mechanical, and it starts read-only. You connect AWS with a read-only role, and ZopNight discovers every SageMaker Endpoints across your regions and accounts. It records a per-action permission verdict for each one, so you can see where it can list a resource but not yet stop it, and you review that inventory, filter it by status or type, and search for the specific resources you care about before anything is scheduled.
Scheduling itself is a cron you write once in plain terms, stop at 7 PM, start at 8 AM on weekdays, pinned to your timezone so the jobs fire at local business hours rather than UTC. A weekly 24-hour grid shows the schedule visually so you catch gaps and overlaps before you save, and an estimate of active versus inactive hours appears before you commit. Resources attach individually or bundle into groups like “dev-cluster” or “staging-db” so a whole environment follows one cadence.
Actions run in dependency order, so a database comes up before the app server that depends on it. When something needs to stay up, an override forces a SageMaker Endpoints ON or OFF for a defined window, carries a reason so teammates understand why it exists, and expires automatically so nothing is left running by accident. If a start or stop fails, ZopNight retries up to three times and falls back to a dead-letter queue rather than silently dropping the action, and every state change lands in an audit trail that records whether a schedule, an override, or a specific user triggered it.
Getting started
Getting started is intentionally low-stakes:
- Connect AWS with a read-only role. Nothing is scheduled or changed at this stage.
- Let ZopNight discover your SageMaker Endpoints and review exactly what it found, filtered by account, region, and status.
- Create a schedule in your timezone and attach the non-production resources or groups you want it to cover.
- Watch the first cycle run, with Slack, Teams, or Google Chat notifications on every start, stop, and failure, then layer in idle cleanup and guided rightsizing.
Production stays excluded by default throughout, and because discovery and recommendations are read-only, you can prove the value before you enable a single action.
Questions we get a lot.
If yours isn't here, email us and we'll answer directly.
How does ZopNight know an endpoint is idle?
It reads the endpoint Invocations metric over a rolling window. An endpoint with no invocations for the window is flagged as idle, with the current cost attached.
Will this affect production inference?
No. Production is excluded by default. Off-hours scheduling applies only to the non-production endpoints you choose.
What other SageMaker objects are covered?
Notebook instances, HyperPod clusters, domains, and Studio apps are discovered too, with dedicated cost calculators and recommendation rules across the SageMaker family.