Design and Evolution of Tencent Cloud Product Metering and Billing System

The presentation introduces the cloud product metering and billing workflow, covering both prepaid and pay‑as‑you‑go models, various account types, and the end‑to‑end transaction process from order creation to resource delivery.

It describes the metering model, including raw data collection (duration, count, traffic), derivation of usage metrics, aggregation at different granularities, and generation of billable data for hourly, daily, and monthly billing cycles.

Multi‑dimensional analysis applications are explained, showing how real‑time queries, usage‑based strategy adjustments, and data‑model examples enable precise usage estimation, resource‑package deduction, and financial reporting.

The evolution of the metering system architecture is outlined across three versions, moving from a Lambda‑architecture with separate streaming and batch layers to a unified data‑lake approach that separates storage and compute, improves scalability, and reduces operational complexity.

Operator orchestration is detailed, illustrating how abstracted operators, configuration‑center JSON definitions, and DAG generation in Flink enable low‑code, real‑time and offline processing for diverse business scenarios.

Solutions for hot‑key challenges, system scaling, and smooth version upgrades (both infrastructure and business logic) are provided, including checkpoint‑based replay, sandbox testing, and gradual rollout strategies.

Finally, future directions are discussed, such as further abstraction of model operators, SaaS‑style service reuse, integration of OLAP/OLTP components, and standardized discount and accounting modules to enhance reliability and reduce operational overhead.