In this article, I will discuss the Saga design pattern. First, we'll address some technical challenges of creating ACID transactions. Then, we'll explore the Saga design pattern and how to implement it.

Database Per Service Pattern

First, let's talk about microservice architecture and the database per service pattern. Basically, each microservice in an application has its own dedicated database. This allows each microservice to manage its data independently, making it easier to develop, deploy, and scale each service without affecting the others. For instance, in an online store, the user service, order service, and inventory service each have their own databases. This isolation improves flexibility, as each service can choose the best database type for its needs and makes the system more resilient to failures.

However, there are some drawbacks to this pattern. Data consistency and distributed transactions are two of them. Since each microservice operates independently, it can lead to potential inconsistencies when updates occur across multiple services. Data duplication across the system can also lead to synchronization issues. Each transaction needs to be distributed across the system. If a failure occurs, rollbacks must happen across the system, leading to another problem: distributed transactions.

What is the Saga Design Pattern?

The Saga design pattern is a method for managing long-running transactions in a microservice architecture by breaking down a complex transaction into a series of smaller, independent steps, each handled by a different service. If any step fails, compensating transactions are executed to undo the changes made by previous steps, ensuring data consistency across all services. This pattern can be implemented through choreography, where each service triggers the next step via events, or orchestration, where a central coordinator directs each step. First let’s talk about Choreography.

At my company, we hold a weekly “Lunch and Learn” event that I really like. It lets us share our experiences and expertise. Recently, during a chat with my colleagues, I got some basic questions about dependency injection (DI). This made me think that it would be a good idea to use one of these sessions to go over DI with the team. Also, I plan to write an article about dependency injection and its best practices. In the article, I'll explain what DI is, how to use it effectively, and what the best practices are.

The circuit breaker pattern stops a service from trying again to call another service when the previous attempts have failed multiple times. It's similar to electrical circuit breakers that automatically cut off the current when there's abnormal activity.

In a distributed environment, calls to remote resources may fail due to reasons such as application exceptions, timeouts, authentication issues, or overloaded systems. Usually, resilient cloud applications automatically fix these issues over time, and the calling application manages these errors using a retry pattern.

However, in some cases, these failures can persist, like when a service is down or systems are consistently overloaded. Excessive retries can create a cascading effect, overloading the same resource and impacting other resources as well. Repeated calls can impact both cost and performance.

At this stage, circuit breaker patterns come into play to address this issue. When the callee retries, it begins to assess the problem. If there's a specific error or if the error count surpasses the limit, circuit patterns activate, breaking the communication between the caller and the callee.

The cloud is vast. Azure docs have around a hundred thousand pages, and AWS is just as big. Other cloud providers are out there too. Each gives you lots of apps, different rules, and dozens of integrations, so creating cloud-native ones has its challenges.

They're not identical, but big providers offer similar solutions. For example, Google Cloud Storage is like AWS S3 and Azure's durable function is similar to AWS step functions. Details and rules can differ, but the main idea of the tech is similar, along with the challenges.

Because problems are alike, solutions are too. We can group similar solutions and make templates for each group. Like a cooking recipe guides you to a tasty dish, these templates can be our guide to perfect solutions.