Get early access
Features · Monitoring · Capacity Planning
Monitoring · Capacity Planning

Days until your disk fills,
not just “used %”.

Trend-based forecasts for disk, memory and CPU. Surfaced as days-until-threshold per resource. Per-host detail; fleet-level rollup. The metric you wish your alert thresholds were measuring.

forecast · days to threshold
db-prod-01 /var 18 days
web-prod-04 / 63 days
web-prod-05 / 142 days
cache-2 memory 91 days
based on 28-day trend
— Why this matters

A static threshold ignores the slope.

01

80% falling is fine. 60% climbing is not.

A percentage tells you where you are. The trend tells you whether to care.

02

"Days until" is the only actionable unit.

It fits in a sprint plan. "Disk at 71%" does not.

— What's included

What is forecast.

Disk per volume
Trend-based, projected to the threshold.
Days-until-threshold
Per resource, per host.
Warning and critical
Two classifications, so a forecast is triaged rather than just displayed.
Fleet rollup
Because buying capacity is a fleet decision, not a host one.
— What it looks like

The slope, not the percentage.

82% 90% threshold
Example view · db-prod-01 /var · 30 days observed, projection crosses 90% in 18 days
— How it works

From a slope to a date.

1

Trend each volume

Usage is tracked per mount, not per host — the full disk is rarely the one you were watching.

2

Project to the threshold

The slope is extrapolated to the point it crosses, which turns a percentage into a number of days.

3

Warn, then escalate

A warning date and a critical date, so the first one arrives while ordering more disk is still routine.

— Works with

Connected capabilities

Metrics & Charts
The trend data the forecasts run on.
→
Provider Integrations
Plan + price metadata for the upgrade suggestion.
→
Reports
Scheduled capacity exports.
→
Smart Alerts
Alert on "days until full" instead of "% used".
→
— FAQ

Questions, answered honestly.

How much history is needed? +
Two weeks gives a reasonable forecast; a month is comfortable; three months catches seasonal patterns.
What about sudden spikes? +
Forecasts assume "next period looks like recent periods". A 10x traffic event will of course invalidate them.
Is there a CLI? +
hostatlas capacity — top-N by predicted pressure, exportable.

Plan ahead. Stop firefighting disk.

Forecasts on every plan; provider-tied suggestions from Starter.

Get early access Back to Monitoring