How Alipay Wallet Optimizes Mobile Traffic and Power with Real‑Time Monitoring

Liu Fukang, an Alipay Wallet development engineer, participated in the evolution of the wallet client framework from version 1.0 to 3.0 and now leads the construction of its monitoring system.

Platform Mobile Client Pain Points

Mobile platforms have limited resources such as CPU, memory, network traffic, and battery. Running applications on these constrained devices creates a complex environment. Android fragmentation across versions and manufacturers adds further compatibility challenges.

The rise of mobile‑first strategies has produced super‑apps (e.g., WeChat, Taobao) that run numerous services on a shared platform, making monitoring an integral part of the platform.

Differences Between Client and Backend Monitoring

Client monitoring must address issues that occur on the user’s device, which cannot be directly accessed and often cannot be reversed; therefore problems must be detected and resolved before they impact the user.

Comprehensive Client Monitoring System

The system consists of three parts: comprehensive diagnosis, quick diagnosis, and dynamic upgrade.

Comprehensive Diagnosis

Metrics are collected from the client and reported to the server. Real‑time alerts are triggered when analysis detects anomalies. The monitoring is multi‑dimensional and flexible.

Data collection runs in an independent process to avoid consuming excessive traffic or power. A configurable collection strategy gathers traffic, battery usage, H5 load speed, RPC calls, startup time, etc., each with its own method.

Quick Diagnosis

Using aspect‑oriented programming (AOP), the system automatically records user interactions, logs diagnostic information, and pushes diagnostic tasks in real time. Detailed logs are woven into the code at compile time, independent of language, platform, or container, and can instrument third‑party JARs.

Dynamic Upgrade

Dynamic upgrades are modular and plugin‑based, allowing incremental updates of specific modules without restarting the client.

Power and Traffic Special Governance

Power consumption is linked to CPU wake‑locks triggered by AlarmManager. Using AOP, the system monitors wake‑lock acquisition, as well as LocationManager, SensorManager, and WifiLock usage, enabling fine‑grained control over battery and network usage.

Source: 云栖社区 (Yunqi Community), original article at http://yq.aliyun.com/articles/126