Which autoscaling approach is most suitable when you want to anticipate demand using historical data and trends, not just current metrics?

Which autoscaling approach is most suitable when you want to anticipate demand using historical data and trends, not just current metrics?

• A. Reactive Autoscaling
• B. Scheduled Autoscaling
• C. Predictive Autoscaling
• D. Auto Autoscaling

Answer: (C)

Forecasting future demand using historical data and trends to drive scaling decisions. This approach, predictive autoscaling, looks at past workload patterns, including trends and seasonality, to forecast what the load will be and scales out or in before the actual demand arrives. Because it anticipates needs rather than reacting to them, it helps maintain responsiveness and avoids the latency that can occur when scaling only after metrics spike. Reactive autoscaling responds to current metrics that exceed thresholds, so scaling happens after a spike starts, which can lead to short periods of under-provisioning. Scheduled autoscaling follows pre-set times and doesn’t adapt to actual workload patterns, so it’s useful for known, regular patterns but can misallocate during unexpected changes. The option described here—using history and trends to predict demand—is the predictive approach.

Forecasting future demand using historical data and trends to drive scaling decisions. This approach, predictive autoscaling, looks at past workload patterns, including trends and seasonality, to forecast what the load will be and scales out or in before the actual demand arrives. Because it anticipates needs rather than reacting to them, it helps maintain responsiveness and avoids the latency that can occur when scaling only after metrics spike.

Reactive autoscaling responds to current metrics that exceed thresholds, so scaling happens after a spike starts, which can lead to short periods of under-provisioning. Scheduled autoscaling follows pre-set times and doesn’t adapt to actual workload patterns, so it’s useful for known, regular patterns but can misallocate during unexpected changes. The option described here—using history and trends to predict demand—is the predictive approach.