For most CIOs (chief information officers), the goal of putting all their organisation's applications, including voice and video, onto one big, simple and intelligent Internet protocol (IP) backbone is a top priority. Many have already done so, and, according to Forrester Research, an IT advisory company, about one-third of all organisations will have a unified IP network by 2010.

This rush to convergence is understandable. Many businesses, notes a recent report from Bearing Point, an analyst firm, currently have a patchwork of networks that are "cobbled together, often resulting in inefficiencies, high costs, inadequate disaster recovery and an inability to deliver new bandwidth-intensive applications". Even if the economic arguments were not so overwhelmingly in favour of standardising on one network, the technical ones – in terms of manageability and ease of use – would be.

But there are some big riders to this otherwise simple proposition. Building a successful, unified infrastructure, it turns out, is a far more complicated, and longer-term, project than many managers are anticipating. Most of the big projects publicised today are still in their early stages.

Basic beginnings

First, the assertion that full convergence is happening today needs clarification. Sometimes, the IP backbone covers only certain applications or geographies, and none so far reaches effectively into the all important area of mobile applications (see box, Going mobile).

Second, many integrated networks have not been designed to support the requirements of the truly unified networks that will emerge over the next several years. At the very least, networks will be required to support low latency applications such as voice, high bandwidth applications, such as video streaming, and resource hungry applications, such as transaction management. They must also be able to differentiate between applications so that their performance can be protected if necessary.

Furthermore, few networks have been modelled to anticipate how usage patterns might change once modern, converged applications are installed – and how these applications affect each other. Many businesses, having switched to IP-only networks, have encountered performance problems.

A third issue is that converging different core services onto the backbone, and even adding in advanced features such as ‘presence', using the session initiation protocol (SIP), is of limited value without upgrading, extending and even rethinking the array of applications and processes that will exploit the technology. "If you see this is a way of ripping out costs, you've missed the point of convergence," says John Wright, general manager, Mobility, BT Retail.

In most cases, enhancements will be handled by the major application providers, such as SAP and Microsoft; but in other areas, extensive upgrading and re-engineering will be necessary. A centralised platform to deliver services may be required.

Leading operators and equipment providers are aware of all of these issues, and are designing networks that are ‘application aware'. These vary in sophistication, in their adoption levels and in their ‘openness'. But none of these technologies are yet used on the ‘open'

Internet, so quality of service of, for example, voice over IP (VoIP), cannot be guaranteed unless it is supported from end-to-end by a network operator.

New standards

One widely adopted standard for improving quality is MPLS – multi-protocol label switching. This method, now widely supported by equipment makers, involves identifying the source of IP packets and attaching a label to them according to agreed priorities. From then on, all routers and switches on the network can prioritise the traffic.

In June this year, Cisco, the leading provider of IP telecoms equipment, introduced a more advanced set of products as part of its AON (application-oriented networking) architecture. This involves installing devices on a network (or upgrading some existing routers) that are capable of rapidly reading and understanding the context of the messages. For example, it might be able to differentiate between a purchase order and an invoice.

Nick Earle, vice president of marketing, planning and operations for Cisco, EMEA, says the innovation will have an "impact on enterprise IT architecture as profound as the migration from mainframe to client/server in the 1980s".

Many service operators and system integrators also address the potential application performance issues. BT, for example, takes a thorough, services-oriented approach based on taking a centralised view of the entire unified network. Its application assured infrastructure (AAI) involves analysing network requirements in advance, building performance models, and then implementing a converged network based on technology from a number of vendors.

IT analyst company Gartner says that service providers could soon be delivering application-level guarantees to more than one-third of their clients. But their challenge, it says, is to turn project work into a regular service.

Perhaps the biggest infrastructure challenge, however, is architectural. How are these new, unified IP-based services and applications to be delivered and integrated – and how is it be done rapidly, flexibly and cost effectively?

The telecoms companies, both mobile and fixed, have the answer, in the form of the IP multimedia subsystem (IMS), which will ultimately work on top of a pure IP infrastructure. IMS is an open platform based on SIP and enables the service providers to develop, deliver and manage IP-based services that are end-user and device aware. To those with an enterprise IT background, it is analogous to the way that web services orchestration platforms can be used to deliver plug in services for the service-oriented architecture.

The IMS is capable of supporting a large number of add-on services, and will be able to manage important functions such as security, identity management and billing. For this reason, the service operators see it as vital – if expensive. Forrester Research recently said that IMS systems could cost operators hundreds of millions of dollars each to build.

Analysys, a telecoms research company, recently forecast that all the major mobile operators would deploy IMS systems within five years – even though it warned that these systems are expensive and risky. Although IMS platforms were originated by mobile operators, however, they are likely to be more widely adopted, with major fixed-line operators also seeing this as a way to gain a competitive edge over rivals. By installing a good IMS, operators will be able to deliver a richer variety of services to their customers, often more quickly and more flexibly than those using alternative approaches.

This raises the question: Will IMS also be important to enterprise customers? "We think so. But it probably won't be called IMS," says Jerry Caron, an analyst with Current Analysis, who foresees the development of similar tools for use by enterprises.

Using the IMS model, he says, it will be possible for enterprises to build a framework for cost effectively delivering IP services, such as VoIP, video-enabled applications and ‘presence'. These networks will be able to link seamlessly into external networks, so in this way, some business will be able to become their own carriers – at least for part of the network.

Application platform

One of the benefits of SIP and the IMS model is that it should prove flexible enough for operators (and enterprises) to add in new application functions – even those have yet to be invented or articulated. IMS has been compared to enterprise resource planning (ERP) systems, in that it could support a huge range of corporate applications.

One example is the use of context and profile data, which is not yet supported by SIP and IMS . "There is a third leg (after signalling and traffic bearing) that's coming into telecommunications – a hidden revolution, and that is context data," says Rick Hull, head of network data and service research for Bell Labs, now part of Lucent Technologies.

Context may involve, for example, giving the messaging service information on user context as well as explicit preferences, possibly by profiling recent phone calls and analysing the user's relationships and history. Without this, the new infrastructure could add more complexity, rather than make it easier to use.

With so much complexity, one question repeatedly arises. How much of this will businesses hand over to service operators, and how much will they do themselves? In theory, it will be possible to build or enhance most next generation network applications in-house, especially using tools such as Microsoft's LiveLink server, and link these to SIP-aware services running on remote networks. But most CIOs are likely to weigh up the risks and rewards and opt for an outsourced service, even if they retain some critical applications in house.