[Based on a post originally appearing at http://asanka.abeysinghe.org/2014/07/api-registry-and-service-registry.html.]
Registry acts as a core component in Service Oriented Architecture (SOA). Early SOA reference architecture named the registry as a service broker for service providers to publish service definitions, allowing service consumers to look up and directly consume the services.
Figure 1: SOA triangle
With the evolution of SOA, Registry started to provide more value, such as lifecycle management, discovery, dependency management, and impact analysis. The Registry became the main management, control and governance point in such architecture. It also became a vital component within the SOA Governance layer of the overall architecture. As a product, the Registry started providing three distinct functionalities – repository, registry and governance framework.
Repository functions to store content such as service artifacts, configuration, and policies, the Registry to advertise them for the consumers to access, governance to build management control and policy-based access over the stored artifacts and to connect people, policies and data.
Changing Role of Registry
With changes and challenges taking place in businesses as well as the technical architecture of the enterprise, the role of the registry has changed. Even while the technical definition of a service is based a standard (e.g. WSDL, WADL), the business definition of a service can vary from organization to organization. Therefore, a customizable definition for a registry artifact became a requirement. For example WSO2 Governance Registry contains RXT (Registry ExTensions) to define services. RXT provides a customizable service definition as well as sets the behavior of an artifact when an artifact is imported and operated in the governance runtime.
When it comes to governance, registry became the core design-time governance controller in the enterprise. Features such as Discovery and UDDi compatibility became nicer to have features than the practical use of them. Having said that, runtime wiring based on environment metadata has emerged as a practical replacement for discovery.
In the modern architecture, Services implement business functionalities of an organization, APIs are interfaces for services that allow consumers to consume business capabilities.
During the last decade, services were developed using various service development standards, programming languages, and frameworks. Services were designed and developed – and funded – in silos in each business/organization unit. This led to duplicate services in the same organization which violates the core SOA concept of reusable shared services. Services are more technically driven, designed and implemented by enterprise developers. If we look at a service catalog, more than 80% of the services perform some kind of a CRUD (Create/Read/Update/Delete) operation. Data Services is another common industry term used to describe the CRUD services. The remaining services implement business logic with the help of business rules and CRUD services. A small remaining portion of utility services exists to provide functionalities such as computations and validations.
Figure 2: Service types
The technically driven nature of the services lead to unhappy consumers. As a result, new service implementations were introduced to the market by duplicating as well as avoiding reusability. Some enterprises started implementing wrapper services in front of the actual services and added an additional burden of maintaining a new service layer.
Emerence of APIs
Complex, rapidly changing business requirements for consumer apps have changed the expectations of the services. Consumers increasingly look at APIs to be:
- JSON based
- Secured with OAuth
- Follow WEB API design
Unfortunately, most of the existing services are not compliant with these expectations leaving a huge gap between the implementation and demand. In a consumer-driven market, APIs that do not meet the demand will not have value and will not survive for long. To meet the demand, technical teams started to write wrapper services in front of the actual services. But this created a huge maintenance issue as well as slowing down the time to go-to-market.
Using APIs as the service interfaces for consumers to invoke service functionality, resolves the issues we identified above, mainly converting technical driven services into business-friendly APIs as well as implementing the common reusable services for the enterprise and meet the demand/expectation of the service consumers. There is much more value proposition from APIs to the enterprise. Lets now look at them in detail.
Figure 3: Service and APIs
A pure API layer might be not enough to cater the demand. A mediation layer might be required based on the gap between the services and API. As I explained earlier APIs are light-weight interfaces that do not represent any implementations. Traditional Facade patterns came to the API architecture as API Facades, providing a solution to this problem. Introducing a mediation layer with the API Facade will take care of protocol switching (transports/message formats) as well as security bridging. In addition, it helps to convert the traditional backend services into modern web-api designs by using techniques such as service chaining (light-weight orchestration). More information can be found in this blog. Having said that, API-ready backend services can be directly exposed through the API management layer as an API without going through any additional mediation layer.
There are two main usages of APIs, internal and external. External APIs create an eco system for a business to expose their functionality to their customers and partners in a consumer friendly, secured and governed manner. Internal APIs again help to resolve the broken SOA pattern and provide reusable common functionalities across business units. This allows business units to promote and “sell” their services across the enterprise and maintain and manage by the business unit itself.
Design-time Implications on the Registry
I hope the information provided above is insightful and has helped you identify the usage of the registry and a clear differentiation between services and APIs. Lets now look at the key discussion points of the service registry/API registry. Looking at a reference architecture will be more helpful to identify the concept.
Figure 4: Service and API registry reference architecture
Service definitions will be defined in the service registry, the service registry will maintain the additional metadata about the services and catalog detailed technical definition of the services. Usually service definitions is defined automatically when services are deployed to the service containers, but in some enterprises this happen as a manual process by importing the service definition from various catalogues or service containers. Once the services are defined the registry will create the dependencies, associations and versions of these services and metadata.
We spoke about having a mediation layer to bring non-API ready services into API ready, proxy services defined in the mediation layer. This will also go as service definition in the service registry.
While technical definitions of the services and proxy services are defined in the service registry, consumers of the APIs require a place to lookup the services. This is where the API registry will come into the picture as the consumer-facing API catalog.
The target audience of the APIs are application developers, hence an API registry requires and supports publishers to cater to their expectations such as social features and the ability to subscribe and get an access token.
From an API governance point of view, API publishers should be able to secure the APIs by providing access control to the APIs and resources inside the APIs. Some APIs might require additional control with entitlement and workflow support to get approval for subscriptions before providing an access token.
We already discussed the functional requirements of the two registries, lets now look at the runtime view. The deployment will depend on the nature of the APIs. There are three categories.
- Internal only APIs
- External only APIs
- Internal and external APIs
To facilitate the first category deployment can combine the service registry and API registry to run in the secured network (LAN). But this will provide two different views for the API consumer and the Service developer. External only APIs require the API Registry to expose externally (in DMZ) and service registry to run in the secured network. Internal and external APIs require an external (public) API registry as well as an internal (private) API registry. The internal API registry can combined with the service registry.
Figure 5: Deployment patterns
I hope the information described above helps you identify the difference between services and APIs as well as the benefits of separately architecting a service registry and a API registry. Having two registries helps fulfill the requirements of the API consumer as well as service developers, in addition, this allows you to decouple services and APIs by having an individual life cycle and versions for each of them.
Vice President of Solutions Architecture