AKF Microservice Splitting Principles: X‑Axis, Y‑Axis, and Z‑Axis Partitioning for Cluster Design

When building a cluster, we must first identify the problems of a single node, such as single‑point failure, limited capacity, and insufficient connections.

To solve these issues, servers are clustered and scaled horizontally, introducing the AKF principle as a microservice design guideline.

Microservice Splitting Principle – AKF

The first issue, single‑point failure, can be mitigated by replication, leading to three concepts: master‑slave, master‑master, and master‑backup, where multiple servers provide read/write services.

In master‑slave mode, the master handles writes while the slave handles reads.

In master‑backup mode, the master provides read/write services, and the backup takes over when the master fails.

X‑Axis Splitting

X‑axis splitting is horizontal replication: running multiple instances of a monolithic system to form a cluster with load balancing, using master‑master, master‑backup, or master‑slave configurations.

Y‑Axis Splitting

When certain functionalities generate heavy read/write traffic while others are idle, the heavy‑traffic data can be isolated into separate servers based on business functions—this is Y‑axis splitting.

Different Redis instances handle different business functions, and to avoid new single‑point failures, Y‑axis nodes can be further X‑axis split.

Z‑Axis Splitting

After X‑ and Y‑axis splitting, if a particular business handled by a Redis instance experiences massive traffic, Z‑axis splitting partitions data requests by source (e.g., Beijing vs. Shanghai), allowing the same business logic to be served by multiple specialized nodes.

AKF Summary

X‑axis splitting: horizontal replication (master‑master, master‑backup, master‑slave).

Y‑axis splitting: business‑oriented partitioning.

Z‑axis splitting: data‑oriented partitioning.