Message Push Monitoring and SLA Practices

Message push occurs on mobile devices daily and is essential for improving app activity, user stickiness, and retention.

Key values include increasing next‑day retention for new users, re‑activating existing users through push reminders, and re‑engaging churned users when push permissions remain enabled.

Background and Pain Points : The message center lacks clear latency standards, creating a gap between business expectations and technical reality. Node‑level latency is unknown, third‑party push channels are treated as black boxes, and code quality issues are hard to detect early.

Monitoring Practice : The team introduced SLA (Service‑Level Agreement) monitoring focused on timeliness and stability. SLA defines measurable commitments such as availability, accuracy, capacity, and latency.

For push services, the primary concerns are timely and reliable delivery, so the monitoring targets these aspects.

Node Splitting : The push workflow is divided into independent, monitorable nodes (e.g., authentication, user lookup, fatigue filter, duplicate filter). This enables comprehensive, gap‑free monitoring of each stage.

Metrics : Important indicators include node latency, blocking volume, and push success rate. Latency is calculated by timestamp differences (e.g., fatigue‑filter latency = T7 – T6). Blocking volume reflects the backlog at each node.

Technical Implementation : Metrics are standardized and isolated from the main push flow using Spring AOP and Spring Event, preventing monitoring code from impacting performance.

Results : The monitoring system quickly identifies latency anomalies, provides node‑level performance data, and guides optimization efforts.

Vendor Push Monitoring : Multiple third‑party channels are monitored for success rate, receipt rate, and click‑through rate. Alerts trigger when metrics deviate from expected ranges. Implementation uses a bounded memory queue to keep vendor monitoring separate from the main flow.

Benefits : Early detection of vendor failures, rapid issue resolution, and performance tuning have doubled throughput during peak periods.

Future Outlook : Plans include expanding monitoring to business‑level push volume, funnel conversion rates, and additional performance indicators to further enhance push effectiveness.

Conclusion : The monitoring rollout has delivered significant gains—clear latency standards, improved throughput, timely vendor issue detection, and overall higher service reliability.