How a Bank Built a Scalable, High‑Availability Redis Cache Platform for Rapid Deployment

Background

Redis, a lightweight in‑memory key‑value store, was introduced in 2015 to accelerate hotspot data access for high‑frequency, high‑concurrency transactions. As the number of Redis instances grew, the organization faced challenges in rapid resource provisioning, standardized deployment, version alignment, and centralized operations, leading to the development of a distributed cache platform.

Objectives

Rapid provisioning – Use the private IaaS layer to orchestrate compute, storage, and network resources for fast, elastic cache service deployment.

Service‑oriented delivery – Provide a unified console for cross‑cloud Redis management, including runtime analysis, slow‑query inspection, memory profiling, statistical dashboards, and routine health checks.

High availability – Integrate Redis Sentinel/Cluster with a three‑availability‑zone (3AZ) deployment to achieve zone‑level disaster recovery and automatic failover.

Architecture

Functional modules

Portal – Shows overall health, alerts, and key monitoring indicators for managed services.

Monitoring – Collects runtime metrics, supports custom monitoring items, threshold adjustments, differentiated alert policies, and historical alert queries.

Operations & Management – Automates routine operations through a standardized UI, reducing manual command errors.

Audit & Log – Records user login counts and operation logs for audit purposes.

Statistics – Summarizes operational data to give a quick view of service health and resource usage.

Backend Management – Handles permission control, scheduled tasks, and integration with related systems.

Technical Support – Provides access to manuals, documentation, and feedback channels.

Key implementations

High availability – Redis Sentinel monitors node health and triggers automatic master‑slave failover; Redis Cluster provides sharding and fault tolerance. Instances are evenly distributed across three physical availability zones, ensuring zone‑level redundancy.

Automated deployment – Users submit a Redis service request via the portal. After administrator approval, an orchestrated workflow provisions compute, storage, and network resources, installs Redis, configures networking, and delivers the service within minutes.

Runtime analysis – Real‑time collection of metrics such as CPU, memory, connections, and slow‑log entries. Metrics are categorized and visualized for risk detection and performance tuning. A health‑score model combines availability, memory usage, connection count, and slow‑log frequency to produce a business‑level service health rating.

Intelligent operations – Integrated instance management (add/remove/restart nodes, command execution), online parameter modification (e.g., maxmemory, maxclients, requirepass) with consistency checks, one‑click master‑slave switch (supporting batch operations and status monitoring), and automated system inspection that generates daily reports on resource usage, capacity risk, and compliance.

Permission control – Implements RBAC. After login, the platform queries the CMDB to retrieve the user’s associated business systems and role information, then assigns permissions scoped to those systems and roles, ensuring users can only operate on authorized resources.

Conclusion

The platform delivers a service‑oriented, highly available Redis caching solution that supports rapid provisioning, centralized monitoring, and automated lifecycle management. Future work will focus on deeper cloud‑native integration to further improve efficiency, reliability, and support the organization’s high‑quality development needs.