Prediction volume × 3 unexpected: scale infrastructure?

Ситуация: ML predictions volume jumped 3× за месяц. SLA latency holding (barely). Engineering wants scale. PM asks: is it sticky?

Recommendations ML service. Cost ~50% infrastructure spend. Sudden growth.

Доступные данные

• serving_log: ts, request_id
• product_changes: deployed_at, area
• usage_growth: ts, dau, mau

Задачи

• Volume growth decomposition: organic / feature / bug.
• Sticky vs spike.
• Scale options: immediate vs delayed.
• Cost projection.

Все кейсы для подготовки → ← Все кейсы