Quelles sont les meilleures pratiques pour migrer d’une architecture monolithique vers une architecture microservices ?

That's a big decision you have to take and make sure you have a reasonable business justification to take this giant step. This should not be done just for the sake of doing. If you are doing it just because everyone wants to ride the microservice wagon. That's the wrong reason. To build a good business case for something of this magnitude, first understand the benefits of architectural modularity. Match those benefits to the issues you are facing with the current system. Finally each and every design has tradeoff, microservice is not immune to that as well. Do your trade off analysis for breaking apart the system.

Microservices architecture becomes essential for two primary reasons: 1. Scalability Concerns: With an increase in users or traffic, monolithic applications often face scaling challenges, particularly in databases which predominantly support only vertical scaling. Transitioning to microservices can effectively address these limitations. 2. Large Development Teams: When the team size grows beyond 20-30 software engineers, working on a single codebase tends to lose efficiency. Microservices, characterized by their independent deployability and minimal interdependencies, offer a robust solution to streamline workflows in larger engineering teams.

If you want to deploy the new features and update the application for staying competitive. If your business aim to reduce the impact of failure and ensure high availability. If you want to reduce the operational costs and optimize the resource utilization. If you want to maintain the code quality and empower developers to work more efficiently. If your answer is "yes" to any of them, microservice is better solution for you because it enable smaller, focused development teams. Each team can take ownership of specific services, leading to faster development, easier maintenance, and the ability to use different technologies suited for each service.

In my experience, this is a crucial step as the decision to embrace microservices architecture solely depends on the business goals. If scalability, performance and maintenance are the ones that outweigh the time and cost of orchestrating development, deployment of microservices, then it's a go.

Before identifying metrics to measure success, the first step is to define the scope of your application. Often times you aren’t modernizing a single application, but rather a set of monolithic systems that are interconnected. Once you understand what you want to modernize, you can then begin defining goals and objectives.

Its completely depends on the organization, which migration strategy they opt. Either Big bang or Incremental Refactoring. Completely rebuild the application as a set of microservices from scratch is time consuming and gradually break down the monolith into microservices over time, often starting with less critical modules will be a good idea for some cases. Create new functionalities or modules as microservices and gradually redirect traffic from the monolith to the microservices is a good idea without taking any downtime and in case of failure and big issue, anytime it can be rollback.

Whether you go with a “big bang” or a strangler fig will depend on your organization’s specific needs. For example, you should consider any important deadlines, contract renewals, or dependent services that could be impacted if you choose a certain approach.

- Adopt an incremental migration strategy. Instead of attempting a complete overhaul at once, migrate one module or functionality at a time. -This approach allows you to manage risks, maintain functionality during the migration, and gradually validate the success of each microservice.

To effectively transition to a microservices architecture, it's advisable to start with smaller application segments that are relatively independent or asynchronous, such as reporting and notification services. These components should be driven by generic domain events. Once these services are stable and the team is adept at handling domain events, it becomes more feasible to migrate larger, more complex components.

Selecting the optimal migration strategy is akin to choosing the right path through a complex terrain. The strangler pattern provides a nuanced approach, allowing for a gradual, minimally disruptive evolution. It's like renovating a building floor by floor while occupants remain undisturbed. Conversely, the lift-and-shift pattern is a swift move, akin to transplanting an entire ecosystem. The key lies in aligning the strategy with your specific needs and resources, ensuring a seamless transition that minimizes disruption, mitigates risks, and aligns with user and stakeholder expectations.

In my opinion, it is often easier to combine parts of step 3 and 4 by including the architecture in your microservice diagrams. Determining your approach (serverless, containers, etc.) early on can influence your approach to service delineation. Are you using NGINX or a managed service? If you’re using containers, where are you storing your docker images? These can all be included in the design phase prior to implementation.

Each microservice, encapsulating specific business logic, exposes a RESTful API. Modules interact with dedicated APIs, abstracting the complexity of database operations. The API layer acts as a gateway, managing requests and routing them to the respective module-specific APIs. Web servers mediate communication between different microservices, ensuring secure and seamless interaction. Kubernetes manages these microservices within pods, allowing for easy scaling and robust deployment. By adopting these practices, organizations can build a microservices ecosystem that promotes modularity, flexibility, and scalability while ensuring efficient communication and management of services.

While designing the services itself, start thinking about how they are going to be deployed. The design phase should not only focus on the actual service design but also the test design and CI ( continuous Integration) so that when you start the development you have a quick feedback loop and fail fast.

- Clearly define service boundaries based on the principles of Domain-Driven Design (DDD). Identify and group related functionalities into microservices based on business domains. -This approach helps ensure that microservices are cohesive and loosely coupled, promoting better maintainability and scalability.

An effective microservice should ideally exhibit the following qualities: 1. Single Responsibility: It should be dedicated to one specific function or responsibility, ensuring focused and efficient performance. 2. Decoupling: It needs to be sufficiently decoupled from other services, allowing for independent development and scaling. 3. Asynchronous and Partition Tolerant: Employing event-driven architecture enhances its ability to operate asynchronously and maintain tolerance to system partitions. 4. Comprehensive Testing: Thorough testing is crucial to ensure reliability and robustness, covering all aspects of functionality, integration, and performance.

Selecting the right tech stack is vital for implementing new microservices: - Opt for a language known for its fast execution. - Choose a lightweight framework to minimize boot time. - Use a database and cache system that aligns with your specific use-case.

Implementation is the moment of truth for your microservices symphony. Tailor your tools and technologies like a bespoke suit, ensuring each microservice's unique requirements are met. The programming language and framework become the threads weaving functionality into the fabric of cloud advantages. Embrace container platforms like Docker or Kubernetes, sculpting a deployment ballet that's both consistent and scalable. Elevate your communication orchestration with a service mesh like Istio or Linkerd—a conductor ensuring seamless, abstracted control over the harmonious dialogue between microservices.

Implementing microservices involves not just coding the services, but also setting up the necessary infrastructure. This includes containerization using technologies like Docker, orchestration with tools like Kubernetes, and setting up CI/CD pipelines for automated testing and deployment. Ensure that each microservice is loosely coupled with others and can be deployed, scaled, and updated independently. It’s also important to standardize development practices across teams to maintain consistency and efficiency in the implementation process.

In a microservices environment, relying solely on logs and metrics is often inadequate for pinpointing production errors. Given that a single user request may traverse dozens or even hundreds of services, implementing distributed tracing becomes crucial. OpenTelemetry represents the state-of-the-art in this domain, offering compatibility with a range of proprietary and open-source tools. Notably, Apache SkyWalking and SigNoiz are comprehensive options in this space, and my experience with them has been positive.

Operating a microservices based system is more complex and expensive. In a monolith you have a small or manageable number of servers/machines to monitor and it's easier to understand the state of the system. With microservices it's complex. You need to invest in distributed logging, monitoring and tracing. Microservices are often deployed using containers which are ephemeral and triaging an issue can get overwhelming. Also the cost of all this monitoring, including logs and traces can be expensive so setting up the right tools and just a sufficient amount of retention is important. Tracing requests become quite important in microservices to determine latency and errors.

Monitoring and optimization crown the migration, ensuring your microservices symphony performs at its zenith. Employ tools like Prometheus or Grafana as vigilant sentinels, collecting and visualizing the heartbeat of metrics and logs. Let them sound an alert symphony at the slightest anomaly, directing attention where it's needed. Trace the intricate dance of requests and transactions with tools like Jaeger or Zipkin, spotlighting the narrative of your microservices' performance. Establish a dynamic feedback loop—A/B testing or canary deployments—as the conductor's baton, allowing you to fine-tune and elevate user experience.

Continuous monitoring and optimization are key in a microservices architecture. Implement robust monitoring and logging to track the performance and health of each microservice. This data helps in quickly identifying and resolving issues, ensuring high availability and reliability. Use the insights gained from monitoring to continuously optimize the architecture. This includes adjusting the size and number of instances, improving code efficiency, and refining inter-service communication. Regular optimization ensures that the architecture remains efficient and cost-effective.

Migrating from monolithic to microservices demands a strategic dance. Prioritize clear business goals, aligning them seamlessly with technical objectives. Choose a migration strategy—be it the nuanced strangler pattern or swift lift-and-shift—tailored to your needs. Craft microservices with domain-driven design, adhering to RESTful APIs and maintaining loose coupling. Implement with precision, leveraging appropriate tools and container platforms. Finally, monitor tirelessly, using tools like Prometheus, and optimize through feedback loops, ensuring a symphony of performance in your microservices architecture.

It's tempting to take this migration as a separate project which might take months to years depending upon complexity. The danger lies in the assumption that while you are building new microservices, your monolith remains static, which is never the case. For a business there will always be new features and functionality which will in built in the existing monolith. So it's important to get your strategy right. One should use patterns like Strangler pattern to move functionality out of monolith gradually and also build any new features in microservices if possible during the process of migration. This can ensure that the migration project is successful.