Today's telcos have serious operational data silo problems. Currently, they respond with highly manual stopgaps that limit their ability to field new services, maintain QoE and reap the full benefits of new technologies such as NFV. Although the YANG model may make things better, a real solution requires deployment of an über-model.
Model alphabet soup
First, though, we need a bit of a primer on information and data models. For two information systems to exchange information they must have a common message format structure. The message format defines how the data that is passed gets its meaning. There is a data model behind the message format -- and it's either implicit or explicit. That is, system designers have in their minds a specific structure of parameters that give the parameters meaning. In order for both sides of the interface to have the same model, there has to be a common way of describing the data model.
When multiple data models and their associated interfaces are combined to make a more complex system, the result is called an information model. What lies at the base of the operational data silo problem are a profusion of non-compatible interfaces and thus a profusion of non-compatible data/information models.
Today there are many different non-compatible interfaces associated with many different description languages. Some examples include: XML, JASON, SNMP MIB, CORBA/IDL, WSDL, graphic symbols and English text.
Overcoming the silo problem
To overcome the silo problem, operators first tried to get a comprehensive set of standard network operations information models and interfaces. Because of different evolutionary paths, different standards organizations and the desire of vendors to innovate and differentiate, this approach proved difficult because of the lengthy cumbersome standardization process. For example, 3GPP SA5, after more than 100 meetings, has only standardized 30% to 40% of the cellular EMS/OSS interface.
In the last few years, a telco-led effort created a joint project in this area between 3GPP SA5 and TM Forum. It resulted in the UIM (Umbrella Information Model), an information model that spans cellular and wireline. The UIM was developed in XML/SOAP.
In 2006, the IETF circulated NETCONF as a way of using XML or JSON to install, manipulate and delete the configuration of network devices. Later YANG was developed by IETF to ease the definition of network device configuration data models to be used in conjunction with NETCONF.
Recently, ETSI NFV ISG has proposed using YANG/NETCONF as the meta-language. NFV ISG has started an information model project with a kick-off scheduled for December 1. In addition, CableLabs is hosting an initial multi-SDO workshop in Louisville, Ky., from January 13-14, 2016, to explore how the industry can converge approaches to information modelling for NFV. We anticipate that this project will result in an extension of the UIM to cover virtualization of the telco core network (network functions performed in a centralized data center).
At the May 2015 NFV World Congress in San Jose, Calif., a spokesperson for the OPNFV group working on YANG said during a presentation that in order to get IETF approval, they had to agree to the use of YANG "with enhancements and extensions." The spokesperson said that the group realized this would mean "no standardization," but that it was their hope that they could get the basic things done in a "standard" version of YANG. Thus, YANG-based modeling may be eventually standardized by IETF, but it will only be the basic model and there will be always vendor specific enhancements and extensions.
The vendor implementation challenge
The bubbling alphabet soup is just one part of the problem. Vendor implementations are more of an issue, and are driven by a vendor's desire to value price, differentiate their products and innovate. These vendor desires yield implementations that are 60% to 70% proprietary.
Given all this, it seems likely that the energy around YANG and NFV will move the industry forward in a very positive direction and will help with a solution to the silo problem. However, by itself it is not enough.
First, it will take some time to complete the current standardization efforts. There are likely to be a large percentage of proprietary extensions. In addition, the industry has a huge investment in legacy systems. Also, there are likely to be new and currently unforeseeable technologies that will emerge with new waves of impacts. Finally, to compete in the emerging telco landscape, operators want to be able to rapidly compose and deploy new services products and the silos stand in the way.
Building the über-model
This leads to the conclusion that what is needed is a model of models. In other words a comprehensive model that can include all the underlying divergent models and translate into and out of them. Caroline Chappell, principal analyst, NFV and Cloud, Heavy Reading, calls this comprehensive data model an "über-model."
In order to build this über-model, the best available starting point has been the UIM. If the ETSI NFV ISG effort to extend the UIM is successful, it will provide an even better starting point. It will provide the basic foundation for an über-model that will also ease the integration of legacy infrastructure into NFV.
But, in order to turn either the UIM (or the ETSI NFV model) into an über-model, the über-model has to be continually extended to encompass the proprietary extensions and implementations that vendors add. Because of software upgrades, new product deployments, etc., the über-model must be able to evolve in real time while the system operates without interruption.
A vendor seeking to provide a product to solve the silo problem, then, must provide a system that contains the über-model. Because of constant changes, the vendor must maintain and update the über-model, thus allowing the rest of the vendor eco-system to continue to develop proprietary systems. It is the foundation that is necessary to support management fabrics and choreography.
The über-model fits into a comprehensive solution to the silo problem and will provide one piece of the foundation necessary for NFV deployment to realize its full potential benefits.
— Mark Cummings, CTO, Orchestral Networks, and Jayanta Dey, CTO Media & Telecom Business Unit, Wipro, special to The New IP