自動擴縮排程：減少可預測流量的冷啟動延遲
Autoscaling schedules: reduce cold-start delays for predictable traffic
自動擴縮排程可在可預測流量到達前提高最低副本數，減少冷啟動延遲。每個環境最多可設定 10 個排程，按每日、每小時或一次性時段設定副本數上下限；時段須至少 5 分鐘且不可重疊。可透過 dashboard、CLI 或 Management API 設定，所有排程共用一個時區。
TL;DR
Autoscaling schedules reduce cold-start delays by raising the minimum number of replicas before predictable traffic arrives. You can set a schedule to recur (e.g., weekday mornings) or to run once (e.g., ahead of a product launch). Configure up to 10 schedules per environment through the dashboard, CLI, or Management API.
Autoscaling adjusts the number of replicas to match incoming traffic. When demand increases, new replicas need time before they can handle requests. This is their cold-start time, which includes allocating compute resources, starting the serving software, and loading the model. Requests that wait during a cold start have longer end-to-end response times.
Autoscaling schedules help avoid cold-start delays by starting additional replicas before traffic arrives. You can set a schedule to recur (e.g., every weekday morning) or to run once (e.g., ahead of a launch).
Schedule autoscaling around demand
Many workloads have predictable traffic. Autoscaling schedules let you change your autoscaling settings at set times, so you don’t need to maintain an external cron job to do it.
Recurring peaks: Raise the minimum number of replicas during work hours and lower it overnight. For example, a meeting-transcription app can scale up before meetings begin and back down after the peak.
Planned events: Schedule a temporary capacity increase for product launches, load tests, or migrations to a new deployment.
✕
Illustrative example: starting additional replicas before an expected traffic surge can help keep end-to-end response times consistent as request volume increases.
How autoscaling schedules work
Each schedule sets a time window and the minimum and maximum number of replicas for that window. The minimum sets the lowest number of replicas the autoscaler keeps running during the window. The maximum limits the total number of replicas the autoscaler can run. When no schedule is active, the environment uses its default settings.
To configure a schedule:
Create up to 10 schedules per environment, all using one shared timezone.
Choose a daily, hourly, or one-time window. Windows can’t overlap and must last at least five minutes.
Use the dashboard, CLI, or Management API.
✕
Two autoscaling schedules for a production environment
Get started
Start replicas before your expected traffic arrives to reduce cold-start delays. Follow the setup guide to configure schedules through the dashboard, CLI, or Management API, or talk with our engineers about your workload.
來源：Baseten 工程博客 · baseten.co