Rule-Driven Automation (3/4) A third and final way to automate delivery I will discuss is rule-driven automation. Rule-driven automation ties together event triggers and actions as the environment evolves from state A to state B when changes are introduced. Rules understand what changes are being delivered when, where (the environment) and how. Rules are driven by events and behavior and are fully transparent. Rules are also behind the simplest and most effective tools employed by users of all levels, from the popular IFTTT to MS Outlook, for automating anything from simple tasks to complex process. Why? Because rules are both easy to implement and to understand. Let’s use again the analogy of software development to make the rule-driven concept clear. It reminds me of my university time when I was working with rule-based systems. At that time, we made the distinction between procedural and logical knowledge. Let me recap and explain both quickly. Procedural knowledge is knowledge about how to perform some task. Examples are how to provision an environment, how to build an application, how to process an order, how to search the Web, etc. Given their architectural design, computers have always been well-suited to store and execute procedures. As discussed before, most early-day programming languages make it easy to encode and execute procedural knowledge, as they have evolved naturally from their associated computational component (computer). Procedural knowledge appears in a computer as sequences of statements in programming languages. Logical knowledge on the other hand is the knowledge of “relationships” between entities. It can relate a product and its components, symptoms and a diagnosis, or relationships between various tasks for example. This sounds familiar looking at application delivery and dependencies between components, relationships between applications, release dependencies etc. Unlike for factual and procedural knowledge, there is no core architectural component within a traditional computer that is well suited to store and use such logical knowledge. Looking in more detail, there are many independent chunks of logical knowledge that are too complex to store easily into a database, and they often lack an implied order of execution. This makes this kind of knowledge ill-suited for straight programming. Logical knowledge seems difficult to encode and maintain using the conventional database and programming tools that have evolved from underlying computer architectures. This is why rule-driven development, expert system shells, rule-based systems using rule engines became popular. Such a system was a kind of virtual environment within a computer that would infer new knowledge based on known factual data and IF-THEN rules, decision trees or other forms of logical knowledge that could be defined. It is clear that building, provisioning, and deploying applications or deploying a release with release dependencies assume a tremendous amount of logical knowledge. This is exactly the reason why deployment is often seen as complex. We want to define a procedural script for something that has too many logical, non-procedural knowledge elements. For this reason, I believe that rule-driven automation for release automation and deployment has a lot of potential. In a rule-driven automation system, matching rules react to the state of the system. The model is a result of how the system reconfigures itself: Rule-driven automation is based on decision trees that are very easy to grasp and model, because they: “Are Simple to understand and interpret.” People are able to understand event triggers and rules after a brief explanation. Rule decision trees can also be displayed graphically in a way that is easy for non-experts to interpret. “Require little data preparation.” A model-based approach requires normalization into a model. Behaviors however can be easily turned into a rule decision tree without much effort. IF a THEN b. “Support full Decoupling.” With the adoption of service-oriented architectures, automation must be decoupled so that it is easy to replace, adapt and scale. “Are Auto scalable, replaceable and reliable.” Decoupled logic can scale and are safer to replace and continuously improve and deploy. “Are Robust.” Resists failure even if its assumptions are somewhat violated by variations in the environment. “Perform well in large or complex environments.” A great amount of decisions can be executed using standard computing resources in reasonable time. “Mirror human decision making more closely than other approaches.” This is useful when modeling human decisions/behavior and makes it suitable for applying machine learning algorithms. The main features of rule-driven automation include: Rules model the world using basic control logic: IF this THEN that. For every rule there is an associated action. Actions can be looped and further broken down into conditions. Rules are loosely coupled and therefore can execute in parallel and en masse without the need to create orchestration logic. Rules are templates and can be reused extensively. Rules can be chained and concurrency controlled. Rules handle complex delivery use cases including decision and transformation. The model is a result of how rules interact with the environment. Models and blueprints can also be used as input, but are not a requirement. Get an early start and try Clarive now. Get your custom cloud instance for free.