Stateless and Stateful Systems

Stateless system

A stateless system maintains no state of what happened in the past and performs its capabilities purely based on the inputs we provide to it.

Examples

A contrived stateless system receives a set of numbers as input, calculates their maximum, and returns it as a result. These inputs are either direct or indirect. Direct inputs are inputs that are included in the request, while indirect inputs are inputs that are potentially received from other systems to fulfill the request.

Imagine a service that calculates the price of a specific product by retrieving its initial price and any currently available discounts from some other services, and then performing the necessary calculations with this data. This service is still stateless.

Stateful systems

Stateful systems are responsible for maintaining and mutating a state. Their results depend on this state.

Example

Imagine a system that stores the ages of all the employees of a company, and we can ask it the maximum age. This system is stateful since the result depends on the employees we register in it.

Some interesting observations

• Stateful systems are beneficial in real life because computers are much more capable than humans of storing and processing data.
• Maintaining state involves additional complexity. For example, we must decide what’s the most efficient way to store and process it, how to perform back-ups, etc.
• As a result, it’s usually wise to create an architecture that contains clear boundaries between stateless components (which perform business capabilities) and stateful components (which handle data).

Benefits of stateless systems over stateful systems

Stateless distributed systems are much easier to design, build and scale, compared to stateful ones.

The main reason for this is that we consider all the nodes (e.g., servers) of a stateless system identical. This makes it a lot easier for us to balance traffic between them, and scale by adding or removing servers.

However, stateful systems present many more challenges. As different nodes can hold different pieces of data, they require additional work. They need to direct traffic to the right place and ensure each instance is in sync with the others.

As a result, some of the course’s examples include stateless systems. However, the most challenging problems we cover in this course mainly concern stateful systems.