IP/MPLS network management--automate or disintegrate!

As a network architect for a telecommunications service provider, your task is to realise the strategic vision laid out for the company at board level. Your plans will determine how the overall aims are achieved: how services are delivered and profits generated

In particular, you are largely responsible for deciding how network operations are managed, how services are provisioned and how the engineering task force configures and maintains the network equipment. You will help define the next-generation OSS architecture--and cope with the challenges and shortfalls of existing manual systems.

The harsh reality is that in the majority of today's carrier networks, there is a gap in thinking between the grand ideas of the boardroom (driven naturally by business criteria) and the practicalities of running and managing a service-oriented infrastructure (driven by pragmatism, experience and available resources).

Most telcos have either already invested in or are seriously considering MPLS-based IP networks with the promise of advanced services. This paper aims to show how their success--perhaps--even their long-term commercial survival--depends on the OSS being an essential part of that business strategy and the design of the network.

What has traditionally been the exclusive domain of engineers is now a crucial component in an IP-based carrier infrastructure--an OSS system where automation is integral and paramount.

1. Are you running to stand still?

OSS systems have not kept pace with the demands of IP networking. That fact has slowed the pace of IP implementation not one jot. This should concern anyone involved in network planning and design.

Firstly, the various processes involved in trying to maintain the network are not only manually run but are, secondly, not integrated with one another. Network operations runs its set of processes, provisioning a different set, network engineering another and security yet another view, leading very often to conflicting interests and overlap in processes, resulting in wasted time and high error rates.

Thirdly, the scale of the task of managing a modern IP network outstrips conventional manual methods. There is a massive increase in complexity, based on an increase in network capability and, most importantly, due to the range of new, converged high value services. There is now a huge diversity of devices from numerous vendors, each with its own vendor operating and management software and command language.

Some form of vendor-specific software productivity tool or scripting application offers a measure of configuration, management and service activation control to each active network element, such as routers and switches. This proprietary software demands specific, expensive engineering knowledge and experience.

Carriers are forced to juggle these disparate proprietary software elements in a necessarily fragmented approach. They must try to manage multiple activities at the infrastructure layer while attempting to plug the gaps in management. There is no true integration--there are far too many discrete "touch points" and changes of ownership across the network to manage with any consistency.

A major appeal of IP-based networking is the new advanced services it supports. As the quantity of services grows, so does the pressure on what has become the fraying fabric of conventional OSS. Eventually the weakest parts will unravel at enormous expense and bringing service quality problems in its wake. And this is the very structure upon which the vast majority of today's carriers are aiming to build and grow the next generation new communications services.

With these three drawbacks, no wonder about 40% of an experienced engineer's time is spent doing what are essentially clerical CLI-based tasks. No wonder up to 50% of manual service configurations fail the first time around and no wonder around 80% of network budget is spent keeping it operational--fire fighting and running to stand still. Perhaps most depressing of all for the network planner is that this is accepted as an unfortunate but unavoidable fact of life. It is not!

It is time for those responsible for service provider strategy and those responsible for implementing that vision to wake up and smell the coffee. Carriers embracing VPN/MPLS networking as the basis for the next generation of their service portfolio, maintain this disconnected approach at their peril.

2. Old-fashioned techniques, old-fashioned results

Service providers worldwide are investing in MPLS VPN platforms so as to offer converged, IP services with CoS differentiation and QoS assurances, backed up by SLA guarantees.

The high-level strategy document is agreed by the board and reaches the networks planners who, after due consideration as to how to proceed, issue tender invitations and begin discussions with the major equipment vendors. The planners and network architects then work out the practicalities and decide processes, standards and protocols.

They purchase the components with a sackful of potential functions--vendor ticks in boxes--but little or no thought is given to how the new network will or, indeed, can be managed effectively, other than relying on vendor-specific tools and the experience of the network professionals. The lifetime costs of maintaining the MPLS network--the bright hope of the carrier's future--and running the converged services it enables, are forgotten in the rush to implement the new equipment and be recognised by the marketplace as offering an advanced portfolio.

Further "downstream" still, the engineering task force implement the processes at the operational level across the actual network. Until recently, this workflow has more or less kept pace with the changes in technology--but not any more.

Until the advent of IP, the engineers had a reasonable chance of keeping up with the configuration changes and event monitoring. Today's customer demand unprecedented flexibility so it quickly becomes impossible to support customers' needs in anything like a timely fashion. At the same time, the margin on services is dependent on delivering them at the lowest cost to the carrier. This simply cannot be done through the traditional manual OSS. The question arises: how will these high-value services be defined, tailored and automated?

"As routers and LAN switches have continued to provide more capabilities over the years, they've also become increasingly complex to manage, resulting in a device configuration nightmare for network operators. To manually configure a device in today's complex networking environment means that a senior-level network engineer must know the device's detailed command-line syntax and semantics, which many times means entering thousands of lines of code. Unfortunately, the end result in today's highly competitive environment is lower operational efficiency, higher costs, reduced network optimisation and shrinking margins."

Debra Curtis, Research Director, Gartner Inc.

3. The Emperor's new clothes At board level very few service providers recognise that the new technology they are embracing as the vanguard of new revenue is, in fact, the organisation's single most unstable asset. The Emperor might promise much, but as he stands at the moment, he has an acute shortage of clothes ...

The explosion in complexity, the rate of change and diffusion of services, the speed at which new services are being introduced and the lack of true integration with existing systems mean that the IP layer is wobbling. It is the only layer in their network still driven by proprietary operating systems, such as Cisco's IOS, Juniper's Junos and the equivalents from Alcatel and Nortel Networks. A non-standard element means maintaining specialised in-house skills that are "single threaded", applying only to a particular set of devices or vendor products silo--it is, of necessity, silo thinking.

Similarly, ask any carrier asked about the deployment of a new product and where their concerns are, and the answer will be: 'How do I activate the service so I can bill for it?" This focus on service is also single-threaded, leading to the creation of application silos in response to new products.

The vendors are constantly developing new features while, at the same time, it is a dynamic environment with a high rate of technological and customer-related change--factors combining to produce a highly unstable environment.

Several additional critical points are often being ignored:

--The investment in multi-service infrastructure means that often more than one service will be delivered over the infrastructure. The infrastructure is "multi-personality", so that introducing an activation platform specifically for MPLS VPNs covers only one product set.

--The on-going management of the platform in these cases is often being ignored. Crucial questions are being forgotten, such as: "How do I deal with changes that are not activiation-related? How do I control and upgrade the platform to support new services? How do I build services that extend beyond one technology layer (such as from VPN to VLAN)?

Effective platform management is not just about activation it is about the holistic and automated management of the whole environment.

The solution lies in automated integrated processes for IP network management, delivered as an essential part of the carrier's strategy.

4. "Abstraction"--hide the complexity Automation is the way to progress from the "old" world where, for instance, circuits were put in place then retained often for many years, to the new age of customer-driven, on-demand, highly flexible services.