AKF Principle for Microservice Splitting: X‑Axis, Y‑Axis, and Z‑Axis Decomposition

When building a cluster, the first step is to identify problems such as single‑point failures, limited capacity, and insufficient connections that exist in a single‑node, single‑instance setup.

To solve these issues, servers are clustered, turning one instance into many. The article introduces the AKF principle—one of the microservice design principles—covering three axes of splitting.

Microservice Splitting Principle AKF

X‑Axis Splitting (horizontal replication) involves creating multiple identical instances of a service and using load balancing. It includes master‑master, master‑slave, and master‑backup replication models, each providing different read/write responsibilities.

In the master‑slave model, the master handles writes while the slave handles reads. In the master‑backup model, the backup does not serve traffic but takes over when the master fails.

Y‑Axis Splitting (business‑oriented partitioning) separates frequently accessed functionalities into dedicated servers, allowing each Redis instance to handle a specific business domain. To avoid new single‑point failures, Y‑axis services can be further split using the X‑axis method.

Z‑Axis Splitting (data‑oriented partitioning) addresses scenarios where a single business service experiences massive data access. By analyzing request sources (e.g., by region), the service can be split further to distribute load.

AKF Summary

• X‑axis: horizontal replication (master‑master, master‑backup, master‑slave) to eliminate single‑point failures.

• Y‑axis: business‑level partitioning to isolate high‑traffic functionalities.

• Z‑axis: data‑level partitioning to handle massive request volumes by further dividing data based on access patterns.