Introducing the Third Network

Date: Thu, 11/13/2014 - 15:53 Source: MEF PR Department

It is time for The Third Network. In September 2014, the MEF introduced The Third Network, which promises Agile, Assured and Orchestrated networking, all delivered worldwide as a service

Introducing the Third Network

MEF meeting in Düsseldorf, Germany. A roundtable discussion on The Third Network

Image credited to MEF

A roundtable discussion on The Third Network, was held on 15 October at a MEF meeting in Düsseldorf, Germany. Participants were:

• Caroline Chappell, Senior Analyst, Heavy Reading


• James Walker, Advisory Director, MEF Board, Founder & President CloudEthernet Forum, and Vice President, Managed Network Services, Tata Communications
• Vinay Saxena, Chief Architect, HP
• Phil Tilley, Sales Marketing Director, Alcatel-Lucent
• Nan Chen, President, MEF
• James Armstrong, Executive Vice President, Spirent Communications
• Ulrich Kohn, Director, Technical Marketing, ADVA Optical Networks
• Nicolas Fischbach, Director of Strategy, Architecture and Innovation, Colt
• Andrew McFazden, Head of Global Marketing, Network Services, Orange Business Services; Chairman, MEF
• Chris Purdy, Chief Technology Officer, CENX
• Gint Atkinson, VP Network Architecture, KVH
• Carsten Rossenhöevel, Co-Chair, EMEA Marketing, MEF & Managing Director, EANTC
• Mark Cohn, Senior Director, Market Development, Ciena

The panel moderator, Caroline Chappell, Senior Analyst for Heavy Reading, started the conversation by asking the MEF’s Nan Chen to define The Third Network.
Chen continued, “Is there a new type of network service which combines the best of both worlds with Internet-like agility and ubiquity with Carrier Ethernet-like service assurance? We believe so. In order to make that happen one of the key enablers is service orchestration. In the MEF we call that lifecycle service orchestration. That really defines the entire lifecycle of services: fulfillment, control, performance, assurance, realization, analytics, and security.”
Why is orchestration so important? Chen explained, “Through the lifecycle we really want to make sure that the services and functions being defined are clearly defined. Today, service providers have to develop their own customized service orchestration. So [The Third Network] enables them to be able to not only communicate vertically within their stack but also vertical and horizontally to communicate with other service providers.”
Andrew McFazden from Orange Business Services described service orchestration, said, “To me it's network management. It's how do we automate a lot of the things which today we do in a manual way, which again is looking at the objective from a service provider's viewpoint . What is it that we are trying to achieve with all of this? And it's agility. We need to do things quicker in a much more lighter way and a much more automated way. And so to me, from a service provider's standpoint, orchestration is all around how you do that, and how do you do it without manual intervention.”
CENX Chief Technology Officer, Chris Purdy, built on that theme: “I have some pretty strong opinions around what service orchestration needs to be versus OSS. Operations Support Systems today in almost all service providers are vertical stacks of functionality and they perform that functionality across all services and all network technologies. That way every OSS needs to know everything about every network [domain]. Service orchestration needs to be done per service [domain] and we need to first start with defining the services.”
Purdy continued, “For example, an Ethernet service is a well-defined entity. And the whole point of a service orchestration solution is that, first of all you, define that service in a completely abstract way that's independent of the underlying technology, and then the orchestration part of it says, okay what are all the lifecycle operations you'd want to perform on that service, make sure you've got the REST APIs and the like. And the orchestration is all about enabling the workflows, the feeds and the analytics to essentially enable that service for its full lifecycle.”
Ciena’s Mark Cohn tied service orchestration to the Open Network Foundation’s Software Defined Networks (SDN), which defines which network activities happen in the control plane and which happen in the management plane. “We need to have the common understanding, and we've already had some discussions over the last year and a half with the ONF and MEF and sat down, and one of the challenges was reaching the common understanding of what is the broader architectural framework that we are working in.”
Cohn continued, “Some degree of overlap is inevitable and it's okay but we try and minimize that overlap and then organizations can decide how the pieces that they are working on are put together. Moving forward with The Third Network is the same thing even though there is less and less overlap when we move higher and higher in the orchestration layer, because the ONF clearly is not.
Gint Atkinson from KVH followed that point. “Does orchestration understand the differences between the implementations of service execution? Should orchestration see that it's been configured through the management plane or through an SDN controller or through a control plane? I think orchestration should be liberated from that level of detail. We definitely need a layer of abstraction there. But more difficult is what does the orchestration engine need to understand in detail about two services or two components so that it can put them together and in a service chain possibly or new service package. And I think that's a difficult existential question right. But definitely we don't want the orchestration to have any idea of how it is actually implemented nothing more than knowing how to discover an interface to make a request for something but I don't think orchestration should really be driving down into management interfaces, control plane interfaces and so forth.”
Hewlett-Packard’s Vinay Saxena jumped in, and said, “What is a service? If you look at services at a very high level that is abstracted from consumer point of view but again you have resources as a services which in some cases means recovery and restoration. So we need to look at services and define what services are. We are starting to see is people are just latching onto these terms of service orchestration saying this is one monolithic box that can do everything.”
MEF’s Chen agreed: “We wanted to have the industry converge on a kind of different layers. Maybe potentially have a four-layer model where the bottom line is potentially the element that is actually infrastructure. And then the next one on top of that could be the control management layer which is controller potentially and then the one on top of that, the service management which actually delivers services, so that could be multiple. And that means a lot, that means connectivity services, that means computing services, that means infrastructure services. And there is one — on top of that is a business management which is actually recording all the applications and things of that nature.”
James Armstrong from Spirent Communications, thinking about implementation of communications and control, said, “We are elaborating on the interface though, the interface is moving from sort of management mindset to API and that means for vulnerability that means we certainly don't need to fix all the different variables that we have today. It could be anything and it shouldn’t change that API if we have a brand new programmable network. And the network is going to expose some broad capabilities at a high level, and then once those capabilities are exposed to applications then we don't have to put constraints unnecessarily helping them to operate.”
CENX’s Purdy replied to Armstrong’s comments: “The biggest challenge with that has been the cost of development. If there are too many variables then it costs you more to build in support for all of those variables. And ultimately when things break, it adds complexity. That flexibility adds complexity, which is hard to work through. So I think the key is, right now as you said, there are so many different variables. And some of them actually do differentiate your servers, like what kind of protection or what kind of delay or delay variation guarantees you offer — those really do differentiate the service. Whereas others — variables like as I said what Ethernet type or what — they don't differentiate your service. And ultimately we as an industry have to standardize, because otherwise it's just adding cost to support all these variables when you don't really need them.”
Phil Tilley from Alcatel-Lucent related The Third Network to Ethernet networks: “We've been deploying Ethernet networks for the last ten years. So are we going to now go and refresh that ten-year cycle to get programmable networks or do we have to work with what's there and say actually the programmability of those networks isn't going to happen tomorrow. The service we want tomorrow or we actually want the service today. It's idealistic to think about the programmable network.”
The panel moderator, Caroline Chappell, replied, “There are examples of operators out there who are automating their networks today. And they themselves are writing their own APIs. They're writing their own scripts, they're writing their own automation. Because they are cloud service providers in many cases, who want to be able to spin up the connectivity with their cloud services.”
James Walker from Tata Communications had strong opinions on that point: “Most people in this industry will struggle to come up with a description of a cloud service. It really means something different in every single circumstance. Certainly it's the case that we've very focused on the end customer. So we accept from day one that the customer is not going to buy a service from a single network service provider. They are not going to buy a service from a single data center service provider. They're not going to buy a service from a single cloud service provider and nor would it be in their interest to do any one of those things. So we have to start off with the assumption that all of us have got to get there and it’s got to be at some level interchangeable.”
Walker continued, “The question of what your service value is to you and your differentiation to me sits in your ability to deliver that piece. So we have starting points scattered all over the place, some of which are connected to each other and some of which are not; some of which are not positioned with each other and some of which are aligned. And there's a huge, huge, huge piece of work that needs to get done to do that.”
Carsten Rossenhöevel from EANTC added, “Maybe we have to live with the situation where you've got standards emerging from proprietary starting points because that's faster. And because that is a prototype testing concept that used to be the idea many decades ago. Like, don't standardize something that hasn't proven its value in at least some prototype situations and maybe we have to turn around and say okay, we are from a manufacturing industry that started MPLS and in industries. Why do we need to fight in ITU on the government layer for A or B, if it just takes two years for nothing? We may as well accept the thing.”
Colt’s Nicholas Fischbach asked, “How much demand is there for the integration of services today? If somebody comes and says let us do it then we just do that. Being in the driving seat with this, it's fighting for something which is completely undefined and probably will need support from the service providers.”
Ulrich Kohn from ADVA Optical Networks, said that while it’s important to define a service lifecycle, it will look different from various perspectives, such as that from a service provider or from a customer. So much, he said, depends on the business case. Is there a business case?, he asked.
The group consensus was that customers and carriers do want the capabilities of The Third Network – quality, ubiquity, and service orchestration, enabled by a programmable network. And that it’s going to require a lot of work, and industry cooperation, to get there.

valorar este articulo:
Your rating: None

Post new comment

Datos Comentario
The content of this field is kept private and will not be shown publicly.
Datos Comentario
Datos Comentario