As the need for IT to absorb variations increases, with the demand for greater business flexibility,we are confronted with some basic questions: how do we design simple for today, to get the current job done, but not "box " ourselves in a corner so we can support the required flexibility ?I think one major answer to this problem is Variation-oriented Analysis and Design (VOAD). VOAD consists of three main types or axes of variation: structural (type|data), process and policy/rule variations. Structural variations are often based on the identification of Types: Customer Type has variations like Gold Customer, PLatinum Customer and Normal Customer. Often this relates to variations in the structure of the class, or of attributes and data associated with the entity (looking from both OO and Data views).Process variations are when you recognize that a Gold Customer may start from a common base, but branch out into a different set of activities for registration or loan processing, for example. Policy or Rule Variations relate to the Rules and Policies (Rules about rules) that apply for each Type of Customer, for example.Clearly these three aspects of VOAD are related and complementary. It is often useful, in practice to distinguish and treat these three axes of variation.Once you analyze the variations along each axis of variation, you come up with a set of variation points: things that will tend to change or remain less stable. Deal with each variation point by applying a pattern. For example if a variation point for calculation of interest is required for various types of Customers, then a Strategy Pattern would be used to handle / instantiate that variation point.Note that variations tend to occur across all layers of an SOA..Not all that changes is a variation that is warranted to capture and model: only those that are architecturally significant will be worth your while to consider. How do you tell? An architecturally significant variation is one in which impacts the architectural /design decisions you will make and have a trickle down effect that will influence subsequent decisions in how you will build your architecture (e.g., SOA). The "domino" that will alter the course of other decisions is a significant or relevant variation.ANother FAQ is why focus on variations? Variations are more difficult to handle than commonalities. Previous literature focused more on identification of commonality which IMO has less of an architectural ripple effect than understanding, isolating and externalizing variations.
What do you think?[Read More]
Data Science, Machine Learning & API / SOA: Insights and Best Practices
From archive: September 2006 X
Ali_Arsanjani 120000D8QB 2,725 Views
After quite a lot time of blog-draught for me, travelling frenzy overtaking me, as I travel to my clients and to share best-practices on SOA with my colleagues around the world, I find that we are indeed reaching a new phase transition in SOA. Last year I saw many look for solutions that called for the first phase of service modeling, namely what I call Identification. The latter part of last year and this year has been more of the next phase: Specification (design of services components and flows). Now we are beginning to see more of Realization of SOA; including prototypes that are expanded and strengthened into gradually more robust and production systems that support service level agreements...[Read More]