Rhema Bytes: A Factory Approach to Service Engineering
When most people think of a factory, the imagery that is conjured is one of mindless repetition, and the generation of large numbers of low-value items. A good example is a nut & bolt factory. In this world, value accrues to the investors from the little profit made on each one of the millions, or billions, of items.
This does not tell the full story of factories. There is another view that most of us do not readily think of. I call this genre, a compositing factory. Good examples are found in the many custom bike shops found across the USA. Many of who arrange engines from Harley Davidson, and kit from other suppliers, into dream-machines especially tailored for their high-end clients.
Both perspectives have one thing in common. In real life, there are the designers that articulate a general template of the “thing”. And there will be the producers that directly replicate the template, or customise it before replication. The nut & bolt represents the former (direct-replication), while the custom bike shop illustrates the latter (customised-replication). There are templates and meta-templates for both bike and nut. The nut template will be driven by permutations on materials, size and strength, whereas the bike template is a composition of an engine, frame, gears, tyres and some other bits.
In SOA architecture and design, we are also concerned with templates (ABB and SBB in TOGAF). Our templates are sometimes abstract, sometimes concrete, sometimes composite, sometimes atomic. Whether as a reference architecture, or a component design, the focus is on a template that solves a generic problem. However, most of the time, these templates are not to be replicated verbatim. Their value is almost always realised in some composition or aggregative context. Some intelligence in application being sine qua non.
For any enterprise, there will be a minimal set of services that must be realised for the organisation to be a meaningful participant in its sector. In addition to these core services, there are others that help to differentiate the organisation. These can be regarded as the macro templates. At the micro level, we find that each genre of service must complete certain tasks in order to deliver meaningful value to clients. Once again there could be differentiation by way of order, algorithm or complement, but by and by there will be a minimal set of tasks, that all must do.
If we apply the mindset of the custom bike shop to our architecture practise, we should see quite a few tools in-house that we can use/reuse. Some that can be bought, and a few that we need to fabricate. I have found that while many enterprises adopt the “reuse-buy-build, respectively” principle, not all evaluate the comparative costs of these options before making a decision. The consequence is that build, and buy, usually outnumber reuse in most organisations. In the cases where there is reuse, existing services are rendered functionally ambiguous to cater for slightly different use cases.
In a previous article, “Rhema Bytes: The Business to SOA Nexus”, it was argued that architecture should seek to create a platform of agnostic services that are well suited to serving the genre of an organisation, rather than the organisation specifically. If one were to decompose an enterprise, top-down, it should be somewhat easier to identify functionality at its most granular level. Top-down decomposition helps identify functionality at the highest level of abstraction. The analysis of each granular functional unit can help determine the comparative value of reusing, building or buying services that provide the required competence.
So, for a new business initiative that delivers services X, Y. and Z. We could ask if there is a Harley Davidson engine that fulfills that X, a Volvo axle (Y1), Saab transmission (Y2), and Toyota electrics (Y3) that deliver Y, and if a component Z5a is truly unique, or needs to be built, alongside Z1..Z3, and Z4c that do not already exist in our catalogue.
Each service, whether bought, built, or reused, is then properly catalogued as to the value it provides, its comparative costing, and what contexts it is to be used in. Such a compendium, built over time, makes it much easier to assemble solutions. Every installment of this approach makes the next assembly simpler and quicker. This is because most unique use cases/scenarios are covered off in the early solutions, and subsequent projects will reveal fewer unseen scenarios.
A lasting benefit of this mindset is that federation and outsourcing is made that much easier, since the templates for the product/service or its composition are predetermined. This means that production and assembly can be separated, and the build and testing are more effectively decoupled. In a previous article, “Rhema Bytes: SOA Services Abstraction” one such model for templating service genres in a SOA is explored. Combining this mindset with the pieces identified in that article should result in a flexible, nimble and responsive “service factory”.