Microservices: Capability model

Microservices is an area that is still evolving. There is no standard or reference architecture for microservices. Some of the architectures available publicly nowadays are from vendors and, obviously, they try to promote their own tools stack.

But, even, do not having an specific standard or reference we can sketch out some guidelines to design and develop microservices.

Capability Model
Image 1. Capability Model (Seen in “Spring 5.0 Microservices – Second Edition)

As we can see, the capability model is main splitted in four different areas:

  • Core capabilities (per microservice).
  • Supporting capabilities.
  • Process and governance capabilities.
  • Infrastructure capabilities.

Core capabilities

Core capabilities are those components generally packaged inside a single microservice. In case of microservices and fat jars approach, everything will be inside the file we are generating.

Service listeners and libraries

This box is referring to the listener and libraries the microservice has in place to accept service requests. The can be HTTP listeners, message listeners or more. There is one exception though, if the microservice is in char only of scheduled tasks, maybe, it does not need listeners.

Storage capability

Microservices can have some king of storage to do properly their task, physical storages like MySQ, MongoDB or Elasticsearch, or in-memory storages, caches or in-memory data grids like Ehcache, Hazelcast or others. There are some different storages but, it does not matter what type of storage is used, this will be owned by the microservice.

Service implementation

This is were the business logic is implemented. It should follow tradicional design approaches like modularization and multi-layered. Different microservices can be implemented in different languages and, as a recommendation, they should be as stateless as possible.

Service endpoint

This box just refers to the external APIs offered by the microservice. Both included, asynchronous and synchronous, been possible to use technologies from REST/JSON to messaging.

Infrastructure capabilities

To deploy our application and for the application to work properly we need some infrastructure and infrastructure management capabilities.

Cloud

For obvious reason, microservice architectures fit more in cloud based environments that in tradicional data center environments. Things like scaling, cost effective management and the cost of the physical infrastructures and the operations make in multiple occasion a cloud solution more cost effective.

We can find different providers like AWS, Azure or IBM Bluemix.

Container runtime

There are multiple options here and, obviously, container solutions are not the only solutions. There are option like virtual machines but, from a resources point of view, the last ones consume more of them. In addition, usually it is much faster to start an instance of a new container than to start a new virtual machine.

Here, we can find technologies like Docker, Rocket and LXD.

Container orchestration

One of the challenges in the microservices world is that the number of instances, containers or virtual machines grows adding complexity, if not making it impossible, manual provisioning and deployments. Here is were containers orchestration tools like Kubernetes, Mesos or Marathon come quite handy, helping us to automatically deploy applications, adjust traffic flows and replicate instance among other.

Supporting capabilities

They are not related with the microservices world but they are essential for supporting large systems.

Service gateway

The service gateway help us with the routing, policy enforcement, a proxy for our services or composing multiple service endpoints. There are some options one of them is the Spring Cloud Zuul gateway.

Software defined load balancer

Our load balancers should be smart enough to be able to manage situations where new instances are added or removed, in general, when there are changes in the topology.

There are a few solutions, one of them is the combination of Ribbon, Eureka and Zuul in Spring Cloud Netflix. A second one can be Marathon Load Balancer.

Central log management

When the number of microservices grow in our system the different operations that before were in one server now are taking place in multiple server. All this servers produce logs and to have them in different machines make quite difficult to debug errors sometimes. For this reason, we should have a centrally-managed log repository. In addition, all the generated logs should have a correlation ID to be able to track an execution easily.

Service discovery

With the amount of services increasing the static service resolution is close to imposible. To support all the new addition, we need a service discovery that can deal with this situation in runtime. One option is Spring Cloud Eureka. A different one, more focus in container discovery is Mesos.

Security service

Monolithic applications were able to manage security themselves but, in a microservices ecosystem we need authentication and token services to allow all the communications flow in our ecosystem.

Spring offers a couple of solution like Spring OAuth or Spring Security, but any single sign-on solution should be good.

Service configuration

As we said int he previous article, configurations should be externalized. It is an interesting choice set in our environments and configuration server. Spring, again, provides Spring Cloud Config but there are some other alternatives.

Ops monitoring

There is need to remember that now, with all this new instances, all of them scaling up and down, environment changes, service dependencies and new deployments going on, one of the most important things it is to monitor our system.  Things like Spring Cloud Netflix Turbine or Hystrix dashboard provide service-level information. There are some other tools that provide end-to-end monitoring like AppDynamic or NewRelic.

Dependency management

It is recommended the use of some dependency management visualization tools to be aware of the system complexity. They will help us to check dependencies among services and to take appropriate design decisions.

Data lake

As we have said before, each microservice should have each own data storage and this should not be shared between different microservices. From a design point of view, this is a great solution but, sometimes, organizations need to create reports or they have some business process that use data from different services. To avoid unnecessary dependencies we can set a data lake. They are like data warehouses where to store raw data without any assumption about how the information is going to be use. In this way, any service that needs information about another service, just needs to go to the data lake to find the data.

On of the things we need to consider in this approach is that we ned to propagate the changes to the data lake to maintain the information in synch, some tools that can help us with this is Spring Cloud Data Flow or Kafka.

Reliable messaging

We want to maximize the decoupling among microservices. The way to do this is to develop them as much reactive as possible. For this reliable messaging system are needed. Tools like RabbitMQ, ActiveMQ or Kafka are good for this purpose.

Process and governance capabilities

Basically, how we put everything together and we survive. We need some processes, tool and guidelines around microservices implementations.

DevOps

One of the keys about using a microservice oriented architecture is been agile, quick deploys, builds, continuous integrations, testing… Here is where a DevOps culture come handy in opposite to the waterfall culture.

Automation tools

Continuous integration, continuous delivery, continuous deployments, test automation, all of them are needed or at least recommended in a microservices environment.

And again, testing, testing, testing. I cannot say how important in this, now that we have our system splitted in microservices the need to use mocking techniques to test, and to be completely confident, we need functional and integration tests.

Container registry

We are going to create containers, in the same way we need a repository to store the artifacts we build, we need a container registry to store our containers. There are some options like Docker Hub, Google Container Repository or Amazon EC2 container registry.

Microservice documentation

Microservices system are based on communication. Communication among microservices, calls to APIs offered by this microservices but, we need to ensure that people that want to use our available APIs can understand how to do it. For this reason is important to have a good API repository:

  • Expose repository via a web browser.
  • Provide easy ways to navigate APIs.
  • Well organized.
  • Possibility to invoke and test the endpoint with examples.

For all of this we can use tools like Swagger or RAML.

Reference architecture and libraries

In an ecosystem like this the need to set standard, reference models, best practices and guidelines on how to implement better microservices are even more important than before. All of this should live as a architecture blueprints, libraries, tools and techniques promoted and enforced by the organizations and the developer teams.

I hope that after this article, we start having a rough idea about how to tackle the implementation of our systems following a microservice approach. In addition, a few tools to start playing with.

Note: Article based on my notes about the book “Spring 5.0 Microservices – Second Edition”. Rajesh R. V

Advertisements
Microservices: Capability model

Twelve-Factor Apps

Cloud computing is one of the most rapidly evolving technologies. It promises many benefits, such as cost advantages, speed, agility, flexibility and elasticity.

But, how do we ensure an application can run seamlessly across multiple providers and take advantage of the different cloud services? This means that the application can work effectively in a cloud environment, and understand and utilize cloud behaviors, such as elasticity, utilization-based charging, fail aware, and so on.

It is important to follow certain factors while developing a cloud-native application. For this purpose, we have The Twelve-Factor App. The Twelve-Factor App is a methodology that describes the characteristics expected in a modern cloud-ready application.

The Twelve Factors

I. Codebase

This factor advices that each application should have a single code base with multiple instances of deployment of the same code base. For example, development, testing and production. The code is typically managed in a VCS (Version Control System) like Git, Subversion or other similar.

II. Dependencies

All applications should bundle their dependencies along with the application bundle, and all of them can be managed with build tools like Maven or Gradle. They will be using files to specify and manage these dependencies and linking them using build artifact repositories.

III. Config

All configurations should be externalized from the code. The code should not change among environments, just the properties in the system should change.

IV. Backing services

All backing services should be accessible through an addressable URL. All services should be reachable through a URL without complex communications requirements.

V. Build, release, run

This factor advocates strong isolation among the build stage, the release stage and the run stage. The build stage refers to compiling and producing binaries by including or assets required. The release stage refers to combining binaries with environments-specific configuration parameters. The run stage refers to running applications on a specific execution environment. This pipeline is unidirectional.

VI. Processes

 The factor suggests that processes should be stateless and share nothing. If the application is stateless, then it is fault tolerant and can be scaled out easily.

VII. Port binding

Applications develop following this methodology should be self-contained or standalone and does not rely on runtime injection of a webserver into the execution environment to create a web-facing service. The web app exports HTTP as a service by binding to a port, and listening to requests coming in on that port.

VIII. Concurrency

This factor states that processes should be designed to scale about by replicating the processes. What it means, just spinning up another identical service instance.

IX. Disposability

This factor advocates to build applications with minimal startup and shutdown times. This will help us in automated deployment environments where we need to bring up and down instances as quickly as possible.

X. Dev/Prod parity

This factor establish the importance of keeping the development and the production environments as close as possible. Maybe to save costs, no the local environments where developers write their code, here they tend to run everything in one machine but, at least, we should have a non-production environments close enough to our production environment.

XI. Logs

This factor advocates for the use of a centralized logging framework to avoid I/Os in the systems locally. This is to prevent bottlenecks due to not fast enough I/Os.

XII. Admin processes

 This factor advices you to target the same release and an identical environment as the long running processes runs to perform admin tasks. The admin consoles should be packaged along with the application code.

I recommend you to read carefully the The Twelve-Factor App page and its different sections.

Twelve-Factor Apps

WildFly Swarm REST example

Microservices are here. They have been here for a long time now and it looks like they are going to be here for a while. Some technologies have born in recent years to help developers to implement their systems following this architectural style. One of these technologies is WildFly Swarm.

Probably, if you are familiar with the Java EE world, you know the WildFly application server. WildFly Swarm is a derivate of this application server but it has been created with microservices architectural style in mind. Taking the definition from the WildFly Sward documentation page:

WildFly Swarm is a project that has taken the WildFly Java Application Server and deconstructed it into fine-grained parts. WildFly Swarm then allows the selective reconstitution of those parts back together with your application to allow building self-contained executable “uberjars”.

It has some other components and functions but, mainly, reading the definition we can understand the concept.

As I have said this technology is here to fill the gap that the old and big application servers cannot cover (This is a huge discussion) in the microservises world. It is comparable with Spring Boot or Payara Micro. With its different flavors and features, all of them cover the same gap.

One of the things that it calls our attention in the definition is the word “uberjars“. Considered a buzzword or maybe not, it is just defining a JAR file that contains a package and all its dependencies. In this case, for example, it is going to be a package that it is going to contain our code plus all the dependencies to run a server with our application just executing the order “java -jar file.jar” in a command line prompt.

After this little explanation, let’s go to build our first WildFly Swarm application. It is going to be a very simple REST application just with a method returning some text. And it is going to look like basically as a tradicional Java EE application. One of the best things of migrating a Java EE 7 application to WildFly Swarm is that the code doesn’t change at all. We can take the old and monolithic code and run it as an uberjar just with a few modifications in the maven pom.xml file.

The first step is to define the correct pom.xml file. In this file we are going to include the proper dependencies to build our REST service and a plugin that it is going to allow us to create our uberjar and/or run our application.

...
<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>org.wildfly.swarm</groupId>
            <artifactId>bom</artifactId>
            <version>${version.wildfly.swarm}</version>
            <scope>import</scope>
            <type>pom</type>
        </dependency>
    </dependencies>
</dependencyManagement>

<dependencies>
    ...
    <!-- Wildfly Swarm Fractions -->
    <dependency>
        <groupId>org.wildfly.swarm</groupId>
        <artifactId>jaxrs</artifactId>
    </dependency>		
</dependencies>

<build>
    <finalName>RestWildFlySwarm</finalName>
    <plugins>
        <plugin>
            <groupId>org.wildfly.swarm</groupId>
            <artifactId>wildfly-swarm-plugin</artifactId>
            <version>${version.wildfly.swarm}</version>
            <executions>
                <execution>
                    <goals>
                        <goal>package</goal>
                    </goals>
                </execution>
            </executions>
        </plugin>
    </plugins>
</build>

The second thing, it is just add the necessary code to the project to implement the REST service. Like I have said before, it is simple Java EE 7 code, I don’t think it need any especial explanation, if you need some additional explanation you have in this blog some articles about REST services and how to implement them in Java EE 7. One of this examples can be found here.

package com.example.rest;

import javax.ws.rs.core.Application;
import javax.ws.rs.ApplicationPath;

@ApplicationPath("/rest")
public class RestApplication extends Application {
}
package com.example.rest;

import javax.ws.rs.Path;
import javax.ws.rs.core.MediaType;
import javax.ws.rs.core.Response;
import javax.ws.rs.GET;
import javax.ws.rs.Produces;

@Path("/messages")
@Produces(MediaType.APPLICATION_JSON)
public class MessagesEndPoint {

  @GET
  public Response getMessage() {
    return Response.ok("This is our first WildFly Swarm App. Awesome!").build();
  }
}

And that’s all we need to add to our first project. Now, we just need to run it.

The first way to do this, it is just running a simple maven command:

mvn wildfly-swarm:run

This first time is going to download a big bunch of dependencies and stuff, and it is going to take some time. Future executions will be much faster.

The second way to execute our application is creating our uberjar file and running it from the console.

mvn package
java -jar RestWildFlySwarm-swarm.jar

You can see the full code for the example here.

See you.

WildFly Swarm REST example