Scheduled Elastic Scaling¶

Forecast-driven scaling policy — instantiates Elastic Capacity Scaling

Pre-positions capacity against a forecast of a known cycle — season, day-part, or scheduled event — so it is already in place when the predictable peak arrives, sized to hold the service target.

Where a reactive loop waits for a signal to cross a line, Scheduled Elastic Scaling acts on the calendar. It uses a forecast of a recurring or announced peak to stage capacity ahead of demand, so the capacity is already present when the surge lands. What makes it this mechanism is that it is pre-emptive and forecast-driven: it deliberately trades some over-provisioning during the run-up for the certainty of being ready, and it sizes that staged capacity backward from the service target it must hold at the known peak — not from any live metric.

Example¶

A grocery chain knows Thanksgiving week cold: the Wednesday before is the year's busiest checkout day, and the rush builds over an hour, not minutes. A reactive trigger would fail here — cashiers can't be conjured as lines form. So the store works from last year's transaction curve plus this year's forecast to schedule extra cashiers and open every self-checkout lane from 10am Wednesday, sized to hold the target "no customer waits more than ≈6 minutes." Capacity is in place before the first wave, then stepped back down Wednesday evening on the same schedule. The store never sensed the peak in real time; it predicted it and was ready.

How it works¶

It is built from a forecast of a predictable cycle — seasonality, a day-part, an announced event — not a live signal.
Capacity is sized backward from the service target that must hold at the forecasted peak.
Capacity is staged ahead of demand, accepting deliberate over-provision during the ramp as the price of readiness.
It fits best where lead time to add capacity exceeds the peak's onset — people, rooms, cold-start-heavy resources.

Tuning parameters¶

Lead margin — how far ahead of the forecasted peak to stage; more margin guarantees readiness but wastes capacity during the run-up.
Forecast horizon & refresh — how far out and how often the forecast is re-cut; a stale schedule pre-stages for demand that has since moved.
Peak percentile — size to the median forecast or a high percentile; a higher percentile buys safety at the cost of more idle capacity.
Blend with reactive — schedule the predictable base and leave a reactive trigger for the residual surprise.

When it helps, and when it misleads¶

Its strength is that it is the only elastic move that works when capacity is slow to add: it beats the lead-time problem by acting before demand arrives rather than chasing it.

It misleads when the forecast is wrong — the peak arrives off-schedule, or a regime shift breaks the historical cycle^[1] — and you pay for staged capacity that demand never came to meet. The classic misuse is a schedule copied from last year and never revised, so the system faithfully pre-stages for a pattern that has drifted. The discipline is to refresh the forecast, size to a defensible percentile rather than the mean, and pair the schedule with a reactive trigger for the unpredictable remainder.

How it implements the components¶

demand_forecast — the forecast of the known cycle or announced event that drives the staging schedule.
service_level_or_performance_target — the target the staged capacity is sized to hold at the peak; the schedule exists to keep it.

It does not sense live load or set trigger levels — that is the Queue-Based Scale Trigger; nor does it actuate the capacity, which is Cloud Autoscaling or Surge Team Activation.

Instantiates: Elastic Capacity Scaling — it supplies the forecast-and-target side, pre-positioning capacity before the predictable peak.
Sibling mechanisms: Queue-Based Scale Trigger · Cloud Autoscaling · Modular Capacity Expansion · Expandable Facility Plan · Flexible Staffing Roster · Just-in-Time Resource Provisioning · Self-Service Capacity Deflection · Supplier Release Contract · Surge Team Activation · Demand-Based Budgeting

Notes¶

Scheduling and reactive triggering are complements, not rivals. The strongest designs schedule the predictable base of demand and leave a live trigger to catch whatever the forecast missed — the schedule shrinks the surprise the trigger has to absorb.

References¶

[1] A scheduled loop assumes the cycle it was fit to keeps repeating (stationarity). When the underlying pattern shifts — a nonstationary demand series — a fixed schedule quietly pre-stages for a peak that has moved, which is why the refresh cadence is a load-bearing dial. ↩