Monday, June 16, 2014

Evolution of industry IT architectures

Each large system has component nodes which connect with each other (for de-centralised architectures) or a central node (in decentralised architectures). Circuit-switching hard-wires the connections making it difficult to change if a new way of connecting components is required for new behaviour required during system evolution. Circuit-switched connections come in many flavours, but fortunately is easy to distinguish from Packet-switching which always enables easy tear-down and re-wiring of connections. Simplistically described message/event oriented connections tend to be packet switched. But words tend to mislead and one must be careful. The same type of architecture can be recursively present for each component among its sub-components. It is only when this kind of packet-switched connectivity is present down till at respective levels the core threads/processes basically switch messages of different aggregations, that one can really saw that the application is truely packet-switched. An IT architecture of this nature that is highly configurable and better suited for evolution, since new functionally can be grafted into the existing network with minor intervention.

With Machine-to-Machine communication and digitization, the flow of data in various industries will move from its current circuit-switched nature to a packet-switched nature. Circuit-switched data is less suitable for evolutionary changes, while Packet-switched data can evolve more easily to deliver the value that Circuit-switched data can provide and more importantly go beyond it.

There are different wireless and wired messaging technologies and protocols at different levels of aggregations. Various industries need to carefully design their respective message stuctures at various levels of these technologies and design the industy IT architectures at that level of aggregation in terms of configurable processing of these messages to achieve industry outcomes at that level. This needs to be done as an extension of the OSI model to higher levels of aggregations. It is only by doing it this properly that the complexity of the processing architecture can be made coarse-grained enough to be viewed explicitly enough to be separately made optimizable based on higher order observations.


In today's world, in the transportation sector, is there any justification for each moving transportation device (car, bus, train, plane) to not publish a stream of its geocodes? Actually every moving object inside the above device publishing its co-ordinates at each level with information being aggregated centrally makes better sense. Most industries have circuit-switched core data items unneccessarily. Packet switching location and other types of data can generate better value in most industries. With M2M and digitization, most industry IT architectures will evolve towards this by default

No comments: