Embedding e-Framework SUMs

I've already posted on using UML sequence diagrams to derive e-Framework Service Usage Models (SUMs). SUMs can be used to model applications in terms of their component services. That includes the business requirements, workflows, implementation constraints and policy decisions are in place for an application, as well as the services themselves and their interfaces.

However, in strict Service Oriented Architecture, the application is not a well-bounded box, sitting on a single server: any number of different services from different domains can be brought together to realise some functionality: the only thing binding these services together is the particular business goal they are realising. We can go even further with this uncoupling of application from service: a service usage model, properly, is just that: a model for the usage of certain services for a particular goal. It need not describe just what a single application does; and it need not exhaustively describe what a single application does. If a business goal only requires some of the functionality of an application, the SUM will model only that much functionality. And since an application can be applied to multiple business problems, there can be multiple SUMs used to describe what a given application does (or will do).

This issue has come up in modelling work that Link Affiliates has been doing around Project Bamboo, and on core SUMs dealing with collections. The e-framework has already defined a SUM for simple collections, with CRUD functionality, and searchable collections, which offer CRUD functionality plus search. The searchable collection SUM includes all the functionality of the simple collection SUM, so the simple collection SUM is embedded in the searchable collection SUM:

The e-framework already has notation for embedding one SUM within another:

And in fact, the embedded SUMs are already in the diagram for the searchable collection: they are the nested rectangles around "Provision {Collection}" and "Manage {Collection}".

Embedding a SUM means that the functionality required is not described in this, but in another SUM. There is a separate SUM intended for managing a collection. That does not mean that the embedded SUM functionality is sourced from another application: the functionality for adding content, searching for content, and managing the content may well be provided by a single system. Then again, it may not: because the SUM presents a service-oriented approach, the functionality is described primarily through services, and the systems they may be provided through are a matter of deployment. But that means that the simple collection SUM, the searchable collection SUM, and the manage collection SUM can all be describing different bundles of functionality of the same system.

Embedding SUMs has been allowed in the e-Framework for quite a while, and has been a handy device to modularise out functionality we don't want to detail, particularly when it is only of secondary importance. Authentication & Authorisation, for instance, are required for most processes in most SUMs; but because SUMs are typically used as thumbnail sketches of functionality, they are often outsourced to an "Identity" SUM.

That modularisation does not mean that the OpenURL SUM shares all its business requirements or design constraints with the Identity SUM. After all, the Identity functionality may reside on a completely different system on the bus. Nor does it mean that every service of the Identity SUM is used by the OpenURL SUM—not even every service exposed to external users. The Identity SUM may offer Authentication, Authorisation, Accounting, Auditing, and Credentials Update, but OpenURL may use only a subset of those exposed services. In fact, the point of embedding the SUM is not to go into the details of which services will be used how from the embedded SUM: embedding the SUM is declining to detail it further, at least in the SUM diagram.

On the other hand, embedding the Identity SUM, as opposed to merely adding individual authentication & authorisation services to the SUM-

—lets us appeal to the embedded SUM for specifics of data models, protocols, implementation, or orchestration, which can also be modularised out of the current SUM.

No comments: