Current State of Enterprise Networks

Enterprise networks as implemented today are largely unchanged from the mid-1990s. The IT application requirements of a typical enterprise have been based largely on client/server data applications from a small set of primary data centers, centered on access to data used in business processes. For example, a typical data access pattern for a manufacturing enterprise includes components such as sales order entry, order status, accounts receivable, service history, and manufacturing and shipping information.

Using Frame Relay or Asynchronous Transfer Mode (ATM), these enterprise networks are based largely on a traditional hub-and-spoke model. Network architects establish a very limited number of hub location(s) at the corporate headquarters (HQ), regional HQ, or major data center, and they connect branch offices to these hubs using point-to-point connections. This type of network is very easy to design and implement, is relatively cost-efficient, and does not present any significant infrastructure management concerns.

Figure 1-1 shows a map of a typical U.S.-based enterprise network, based around its corporate HQ location in San Francisco. Note that almost every connection from a field office connects to the corporate HQ location, even when another backbone site is nearer.

Figure 1-1. Typical Hub-and-Spoke Enterprise Network

A drawback of the hub-and-spoke model is its inability to adapt to changing application requirements. As business trends shift, global enterprises are looking to increase their productivity, hoping to drive revenues up while keeping operational expenses down. Enterprise IT departments are expected to deliver many of the enhancements needed to drive productivity growth. For the network architect, this shift drives many changes to the enterprise network.

The typical enterprise IT department once approached different voice, video, and data delivery from separate angles. Voice networks were typically built using private branch exchange (PBX) nodes connected to the public switched telephone network (PSTN). Videoconferencing networks were also built on PSTN technology, using Integrated Services Digital Network (ISDN) BRI. Video delivery networks, where used, were built with satellite delivery mechanisms. Typically, each of these had separate IT organizations engineering and maintaining these services.

In an effort to drive down costs, enterprises have been calling for convergence of these technologies, asking for a single communications network that can carry all application types, whether voice, video, or data. Although this can be accomplished in a traditional hub-and-spoke network, other factors are driving a shift in network architecture.

The desire for increased productivity is changing how the typical employee performs his or her job. The use of collaboration technologies to bring together a diverse and dispersed global organization is strongly influencing the development of the enterprise communications network.

For example, a typical enterprise sales force may have once relied only on a top-down sales strategy and local expertise and knowledge to address customers' requirements. This approach is inherently inefficient, because global teams may not be able to collaborate on a unified global strategy. Also, employees with expertise in each region may not be able to regularly share their experiences and knowledge with each other. These increased requirements for collaboration and knowledge sharing have driven a significant number of applications, both asynchronous and real-time, to integrate these groups more tightly as a global pool of resources. This places "real-time" importance on the network: The enterprise network architect must ensure that the network latency between two locations is optimized for this new suite of applications.

Finally, enterprises are recognizing a change in application distribution between data centers. The resiliency in today's enterprise network is an absolute requirement to guarantee the productivity levels in case of a major (or even minor) failure of any component in the network. In the traditional hub-and-spoke network, this typically means that the addition of another data center requires that the number of permanent virtual circuits (PVCs) in the network be doubled.