GLORIAD newsletter May 2007 issue - dvNOC

GLORIAD Developing the dvNOC: Distributed Virtual Network Operations Center

International networking has evolved into an essential piece of the infrastructure supporting

global R&E cooperation. It is both “conveyed by” and a causative agent in the globalization of science. The degree and level of integration of international science and education collaboration will only accelerate in the years ahead. The demands upon the R&E infrastructure – and most fundamentally, the underlying network – will only increase.

Hybrid Services

The necessity of providing a set of hybrid network services is becoming ever more important. While a general routed infrastructure providing simple Layer 3 services will continue to meet the needs of the vast majority of the R&E community, there is a growing set of users, communities and applications that require specialized services that we are trying to provide with Layer 1 and Layer 2 infrastructure managed by the international GLORIAD team and its partners throughout the GLIF community. For many in this growing community the need is for high-capacity, highly reliable, end-to-end dedicated services for specific durations of time – whether for the transfer of massive amounts of data or the control of remote instrumentation or the delivery of quality video or the development of a better integrated supercomputing fabric. These needs are simply not met by general, shared Layer 3 services. If programs such as GLORIAD do not provide such services, the scientists and engineers will have to build – at great expense and with limited network expertise – these same networks. Hybrid architecture for the evolving R&E network brings both a rich set of services but an equally rich set of management problems. The biggest

challenge lies not in the technology but in the management model. When the global R&E network was a very small set of international exchange points with a limited number of routers and relatively simple exchange policies, management of global networking was manageable. With the proliferation of network circuits, equipment, countries and national players, the management of Layer 3 networking globally has become more complex. With a growing set of customers needing specialized, dedicated, end-to-end services, existing models of human and technical infrastructure will not be sustainable.

“Where we actually manage to get things done primarily results because of the “hero efforts” of the network engineers.”

Changing Needs

Centralized management might well reap benefits for comparatively smaller national NOCs, but the sheer size and number of players involved makes this model unsustainable for international networking, whose main characteristic is no central authority for building, managing and growing the networks. And while national network operators do cooperate with other national network operators on many levels, without an established cooperative management model, high levels of visibility into internal networks for troubleshooting and other management tasks are not permitted, and control of network equipment is certainly not afforded to other national operators. Where we actually manage to get things done primarily results because of the (expensive)

“hero efforts” of the network engineers. Establishing and managing specialized services that cross one or more international boundaries has already presented difficult challenges and these challenges will continue as global networking becomes ever more complex.

Three Scenarios

1) Continuing the existing decentralized model based on best practices and hero efforts. Given the increasing complexity of the underlying infrastructure and the increasing demand for both advanced and basic network services, this scenario will inevitably have to transition into one of the other two.

2) A centralized and controlled model of international network management. This is being attempted by major national and international networking efforts. The ongoing consolidation of

these various efforts suggests a more monolithic approach toward global network management. This has as one of its fundamental advantages the simplicity of centralized control and may yield greater network service reliability in the short term. There are many risks to this approach, however, among them that centralized efforts tend to limit and at times even stifle innovation, be it individual, regional or national in scope. The “we will build it for you” idea is appealing on some level, but innovators and the high-end applications that are on the leading edge of discovery and drive the future of networking are better served by a model that can by its nature be highly customized.

3) A collaborative distributed model that is characterized by cooperation among national and regional network operators who insist on maintaining their autonomy and control within their domestic sphere but who wish to contribute to the reliable management of international links and services – because doing so is in the interests of their service constituency. This model suggests agreement on certain operational standards and tools, the very things that are provided by a distributed virtual NOC.

The Future of Global Networking

Through its existing program, GLORIAD has already laid down a foundation toward establishing a distributed managed operating system for a reasonably large subset of the global community: We have cooperatively negotiated contracts for circuits and services and we have agreed with a set of trusted international partners to cooperatively share and manage the services for the benefit of our research communities. The next step is now evident – both for our GLORIAD program, and more importantly, the broader R&E networking community.

“What we point to with the dvNOC is something even larger – and that is the model of international networking dominated by no one, contributed to by everyone.”

GLORIAD already has a large set of trusted partners working toward this same end. The hope is that we can begin to further refine and develop this cooperative, shared model within the GLORIAD framework and develop a set of tools, technologies, frameworks, standards, and approaches that work for us, while simultaneously building a system that is extensible to a global set of contributors. By working initially and directly with the GLIF community we can expand the effort more broadly. As we make progress and as we further open the effort to others, we can shift the direction of global networking toward a global network infrastructure that exists for and is managed by the global R&E community rather than a model dominated by the interests of a small number of large global players (GLORIAD among them). In a very real sense, GLORIAD was created to promote this model of collaborative international networking in which international partners allow certain visibility into each other’s networks, share a certain amount of control, and cooperate for the good of their service constituency (research and education communities). Even for such a project with limited scope as GLORIAD (eleven national partners), shared management of networking is not an easy problem. What we point to with the dvNOC, however, is something even larger – and that is the model of international networking dominated by no one, contributed to by everyone. Such a model of collaborative, reliable operation of network services depends upon a base set of common agreements, adoption of certain standards, tools and technologies – but in no way restricts any national or regional contributors in the use of their own technologies or in the pursuit of their own unique technical agendas.

Our thanks to Dr. Dongkyun Kim of Korea/KISTI for his hard work in helping define first steps for implementing the dvNOC during his stay with us in Tennessee in 2006-2007.

[Image taken from Paul Baran’s RAND Memorandum, "On Distributed Communications: 1. Introduction to Distributed Communications Network" (August 1964)