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

When building a cluster, the author first identifies common problems of a single node such as single‑point failure, limited capacity, and insufficient connections.

To solve these, the article introduces the AKF principle, a microservice design guideline that uses three axes of splitting.

Microservice Splitting Principle – AKF

X‑axis splitting (horizontal replication) involves running multiple instances of a monolithic service, forming a cluster with load balancing. Typical patterns include master‑master, master‑slave, and master‑backup replication. Images illustrate these configurations:

In a master‑slave setup, the master handles writes while the slave serves reads. In a master‑backup configuration, the backup takes over when the master fails.

Y‑axis splitting

This axis separates services based on business functionality. When certain features receive heavy traffic, they are isolated onto dedicated servers, allowing independent scaling. The author notes that Y‑axis splitting can still suffer from single‑point failures, so it can be combined with X‑axis replication for each functional node.

Z‑axis splitting

After applying X‑ and Y‑axis splits, the Z‑axis focuses on data‑level partitioning. If a particular Redis instance handling a business experiences massive traffic, the traffic can be further divided by geographic source (e.g., Beijing vs. Shanghai) to spread the load across multiple identical services.

In summary, the AKF principle recommends:

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

Y‑axis: functional/business splitting.

Z‑axis: data‑centric splitting, often combined with geographic segmentation.

These strategies together improve availability, scalability, and performance of backend services.