The Web services OSS: Adding fundamental value to IP services

In the popular press, the Internet is associated with dotcom failures and market hubris, but the inescapable fact is that networks based on ubiquitous IP transport have become fundamental to the success of national and international commerce. Unfortunately, the so-called "success-disaster" of the Internet is that while convergence of the enterprise applications infrastructure on a unified IP-based network layer has allowed tremendous innovation and dramatic cost reductions, the opposite is true for the service provider.

Our industry is at a point of inflection, one at which a fundamental shift in the balance of power between the enterprise and the provider leaves us wondering what the new provider landscape will be once we've gotten through the current brouhaha over corporate malfeasance and the providers have digested their spree of overindulgent capital investment. At the very least, the notion of IP as the ubiquitous, all-encompassing service network must be questioned. As Dave Passmore of the Burton Group recently pointed out in a feature in Business Communications Review, 50% of the U.S. national IP backbone infrastructure is now subject to Chapter 11 proceedings. Any enterprise that depends on IP-based communications must now urgently assess its strategic options while the carrier market returns to health. More importantly, the development of a stable, profitable IP services market that meets business requirements and allows the service provider industry to rebound requires a rethinking of our current models of services and the process of service delivery. Key to the successful delivery of IP services will be an automated supply chain for the delivery of communications capacity, a new category of software we call the web services operations support system (OSS).

At least two important events-namely, the 1996 deregulation of the U.S. telecom industry and the dotcom boom, in combination with a dramatic oversupply of money from a strong U.S. economy combined to form the tsunami that has led to the dramatic collapse of the provider industry. In the wireless sector, government-led auctions of spectrum for 3G wireless networks, held in a red-hot market for traditional 2G wireless gear, crippled major providers, leaving the industry incapable of building the networks they planned. In the wireline market, providers over-invested by well over $50 billion between 1999 and 2000. Driven by their own marketing hype and by false accounting for network growth using capacity swaps, the provisioning of excess capacity in anticipation of a massive growth in IP traffic led to massive debt and subsequent failure of many providers. 

However, it was not so much the failure of IP traffic growth that caused the problems: Recent research published by BCR indicates that traffic volumes have continued to increase multiplicatively, and that with capex now under tight rein, there is a serious threat of under-capacity in the Internet that will emerge in the fourth quarter of this year. The problem was, and still is, more fundamental than that. Though RHK projects Internet bandwidth demand will grow by about 80% in 2002, the firm also projects a decline in margins per bit of 46%--clearly demonstrating that IP presents providers with a problem that is not demand-oriented, but value-based. The lack of a sound value proposition for IP-based services will inevitably lead to a fundamental reshaping of the telecommunications sector.

The growing portion of ILEC revenue being realized from wholesale services hints at one possible outcome of the process: a trend toward development of niche- or sector-focused vertically integrated service providers that base their premium IP-based service offerings on the wholesale packet transport of the facilities-based providers we know today. In this model, the facilities-based provider simply offers basic Layer 1 and Layer 2 services, with little differentiation other than the basic service attributes required by any of the verticals, namely bandwidth and service guarantees. The largely non-facilities-based vertical service providers implement services at Layers 3 and above, catering to specific, high-margin markets which require application-specific expertise beyond the capability of the existing providers to develop.

For example, consider the evolution of a service provider offering extraordinary security, bandwidth on demand and storage to the banking sector, for offsite disaster recovery of mission-critical data. Development of such vertical expertise is beyond the capacity of today's facilities-based carrier, whether ILEC or IXC, but there can be little doubt that a service of that nature, based on ubiquitous packet-based transport, could command high margins and be very profitable. Another example might be an enterprise-focused content-delivery network with the capacity to outsource the content management of documents and media for large corporations, based on worldwide overlay networks of Akamai-like hosting functionality.

What might drive the development of this two-tier market? First, the ready availability of bandwidth has already commoditized the basic bit-shifting business of transmission and packet transit. Contrary to popular belief, the cost structure within facilities-based providers is dominated by operational expenditure rather than capex-based network deployment. Large spending on sales and marketing retail services can be cut by a move to a wholesale-based service model, in which the retail provider assumes the burden of meeting the market need (and consequently filling the pipes), while the wholesale provider focuses on running a smaller set of efficient pipe services. More importantly perhaps, facilities-based providers today dedicate the dominant proportion of operations expense to the task of managing the tremendous complexity and complex configuration of their multi-vendor, multi-technology networks. As they implement new OSS to automate the complexity associated with service delivery (and in particular the automation of the wholesale/retail provider interface) they stand to reap not only the benefits of cost reduction but also to facilitate the delivery of new service offerings that are impossible to realize today.

How can providers move toward an "automated services supply chain" in which the need for manual intervention at every step of the way is slowly eliminated? Every service provider lists "customer self-service" portals and customer network management capabilities as fundamental needs to remain competitive. Perhaps a more reasonable way to phrase the question is to look at it from the customer side: Why is it so difficult for providers to give their customers the ability to control their own services, to innovate and to modify their needs independent of the underlying network?

Traditionally, providers have built one network per service technology (i.e., ATM, frame relay, Metro Ethernet). But the dramatic pace of technology development has left them with a hodge-podge of networks and equipment from dozens of vendors, each with unique capabilities and limitations. As a result, providers have ended up with massive duplication of systems and personnel to manage geographic distribution, diversity, complexity and scale. Not only is this costly, it makes services inflexible and vulnerable to error-prone manual procedures.

Crucially, there is no dynamic link between the customer's service needs and the underlying network resources dedicated to serving them, making it extremely difficult to relate network performance to customer SLAs or to offer valuable new services that allow customer self-management and even "on-demand" control of service parameters. Today's networks are manually operated, leading to massive inefficiency and inability to innovate.

To tackle the automation problem requires a new degree of sophistication in OSS. The first generation of service management systems has (at best) given providers the functionality of a multivendor element management system (EMS), allowing a human operator to use a single interface to configure multiple devices. Yet there is no link between what the customer wants and how the network delivers it. Template-based configuration systems allow a crude service model to be represented as a standard "form" that pushes device-specific parameters to the network, but such systems operate without knowledge of the currently configured services or available resource in the network.

Figure 1: The Evolution of Automation and Control

The same is true of so-called policy-based network management (PBNM), which uses a rules-driven approach to configuration. Because policies don't take existing service commitments into consideration, PBNM can easily lead to SLA violations for customers already on the network as new services are turned up.

To achieve the level of sophistication in which the industry can once again begin to innovate and find new revenue will take a new category of software that is capable of intelligent automation of the most critical, expensive and error-prone processes in the delivery of rich services on multitechnology networks. Crucially, this software must incorporate a highly detailed model of the network being managed, its topology, its capability to deliver various services and its current commitments to existing customers. And it must be able to assign the network resources in a way that best meets the needs of each customer while maximizing the revenue potential of the network as a whole. 

This emerging category of software, known as service control software, is based upon two key components that, combined, have the potential to deliver on the vision of a cost-efficient network with an automated supply chain for network services. The components tightly connect customers, their services and the network resources that serve them, ensuring customer satisfaction while putting network capacity to best use.

Service abstraction

Both wholesale and retail providers must be able rapidly innovate in communications-centric services without having to focus on the underlying complexity of the network. Separation of the semantic concepts of the service from the underlying network capability that delivers them dramatically accelerates service development and facilitates competition. To achieve an automated customer/provider interface requires the ability to capture customer-specific service parameters in an XML-based service level specification that defines "what" the customer requires from the network but not "how" to deliver it. The SLS allows service parameters to be changed on demand, making possible a new realm of services such as "bandwidth-on-demand" that require dynamic reconfiguration of the network. Use of an XML/SOAP-driven activation process aligns naturally with the emerging Web services framework, and is key to the automation of the "network services supply chain". More fundamentally, it allows services to evolve rapidly because they are independent of the network, and the network can evolve without requiring any changes to the abstracted service logic.

Network automation

On the wholesale provider side of the customer/provider interface, an automated translation of the SLS into the required configuration of the network is a necessity. Service control software systems now emerging on the market are able to model the physical and logical topology of the network in detail, automatically gathering information from the network to reconcile the intended and actual states of the network, and tracking demand for services and capacity. Based on this detailed understanding, the delivery of services can be tightly controlled, putting the provider in position to implement granular control over their infrastructure with functions such as admission control, to decide whether or not new service demands can be met without compromising existing service SLAs; traffic shaping and class-of-service marking, to guarantee that the customer receives an agreed bandwidth and service level, (e.g., premium, standard, best effort.); and bandwidth management, to share bandwidth between different services so that each can meet its SLA. These functions ensure that the network resources meet the needs of the individual subscribers and their services at all times, as well as delivering on the service provider's need to maximize revenue from the network.

Finally, the network must be configured automatically, but under full control of the operator. When a service is being implemented in the network, service parameters must be translated into vendor-specific configuration actions as a single transaction, hiding the complexities of protocol adaptation and vendor-specific feature implementations from the service layer. The operator must be in full control of the process, with the ability to check any proposed changes, ensuring network stability and integrity. If the configuration fails, the network must be automatically rolled back to its previous configuration.

The emerging category of service control software addresses the key issues that make service automation a nightmare today, namely an inability to deal with massive scalability, diversity and network complexity. By linking customers to their service and also to the network resources that serve them, this new category of OSS is enabling providers to achieve enhanced control over their networks, while cutting the cost of operations. It is also key to the development of a value-added IP services model, one in which profitable services developed over a differentiated, tightly controlled network infrastructure begin can address the demand for value-added packet-based services. Crucially, service is separated from the network, allowing the provider to maximize the value of their existing multi-technology, multi-vendor network.

Simon Crosby is Founder and Chief Strategy Officer of CPlane Inc.

Visit CPlane online.