As NFV and SDN exit trials for production, the industry is acknowledging that the idea of "The Network" is mythical. Even within a single operator, it's not "The Network," but many networks -- an archipelago of operational silos run by humans with slow rowboats -- and existing information models won't provide the necessary speed to bridge the silos.
Today, in order for services to run at all, network operations are run by human beings -- expensive expert technicians slowly rowing information from one island, or silo, to another while cleaning, shaping and translating the information by hand. If communications service providers (CSPs) so desperately need to fulfill their ambitious agendas for increased agility and reduced costs, these slow rowboats must be replaced by bridges -- and fast.
The challenge of bridging the gap between the demand for high-automation and hindrance of operational silos is gaining recognition as evidenced by developments like the Metro Ethernet Forumís (MEF) Lifecycle Service Orchestration (LSO). However, the LSO vision is predicated on a standard, shared common information model.
Although more up-to-date standards for universal information models certainly have an important part to play, better models alone will simply not solve the problem. They will take too long to standardize, and in any case, as history has expensively demonstrated time and time again, they cannot account for the huge degree of variation across service providers and equipment vendors without compromising the competitive advantages embodied in their different approaches to service design and the challenges of network operations.
The unacknowledged conflict between the need for interoperability and the imperative for both providers and vendors to differentiate themselves in the market place has always existed. However, the idea that these technologies will eventually cut this Gordian Knot by decoupling the definitions of services from the hardware and operational processes that deliver those services is a seductive aspect of the messaging around network functions virtualization (NFV) and software-defined networking (SDN).
There may well come a time in the future when the network is entirely commoditized with all the value confined to abstract service definitions. However, given both the economics of the investments in current infrastructure, and the operational realities evident during any engagement with a network operations department, this is by no means an imminent state of affairs. Plus, service providers canít afford to ignore the reality of operating profitably while they wait for it to come about.
So historically -- and probably economically, too -- across-the-board standardization of models (and therefore, by implication, interfaces) canít provide pre-fabricated bridges between the operational islands and building every bridge to measure is just too expensive.
But we aren't out of options because there's data, and lots of it.
Elsewhere in the vastness of the tech universe, huge changes have been taking place in almost every aspect of the way that data is represented, stored, queried and processed.
Gone are the days when an information model was forced to be complete in every respect -- and thus entirely inflexible. Today, data technology exists that makes adaptation cheap, fast and easy, but this isnít yet being factored into the way that we, as an industry and as a technical community, think about data and information models.
We still seem to crave the security of hefty tomes filled with standards-body-approved UML without acknowledging that this very level of detail transforms the standard model from useful best practice into a millstone of future cost and risk. With some exceptions in measurement and performance management, the overwhelming majority of OSS, BSS and network management software currently deployed predates advances in available technology, so itís understandable that the existing landscape looks the way it does.
However, there is an urgent need for a broader discussion about the nature and role of common information models in efforts such as the MEFís LSO that are aimed at a future in which interoperability without human intervention will be mandatory.
The limiting factor in the scale of adoption of NFV will be the ability to bridge the archipelago of domains -- to automate service lifecycles across multiple NFV domains and "classical" domains of existing infrastructure. And the single biggest obstacle to this is data interoperability.
Because each island in the archipelago speaks its own language, to support the ultimate goals of CSPs and deliver the degree of service agility required to satisfy the business case for NFV, the model of translation between these languages needs to be simultaneous.
The bad news is that this cannot happen using the approach to information models that has been prevalent for the last 20 years. The good news is that data technologies now exist that enable new approaches -- but we as an industry need to embrace these and have the courage to be radical in the famously conservative world of standards.
— Leo Zancani, CTO, Ontology, special to The New IP