One constant in the IT industry is rapid change. One important way that organizations can effectively manage change is by developing and maintaining an enterprise network architecture. A network architecture ties an organization’s business goals into its network planning process in order to create a strategic direction (or road-map) of how the network will evolve over the next three to five years based on long-term business objectives. Key to an effective network architecture is the ability to understand what new technologies are emerging that can be leveraged to provide a competitive advantage.

In this article, we peek into our crystal ball to try and predict the most significant emerging technologies that will impact network architectures over the next two to three years in three crucial areas, Wide Area Networking, Convergence, and Quality of Service.

The Future of the WAN

WAN Access

For the most part, access to the WAN is still based on time-division-multiplexing or Frame Relay, providing access at speeds from 56kbps (DS0) to 45Mbps (DS3) per second. In many areas, ATM access services are available at speeds from DS3 to OC3 (155 Mbps). Incumbent exchange carriers still tightly control access in most areas and the competition that was envisioned as a result of the 1996 telecommunications reform act has been slow to materialize. However, within the last year the signs of competition are beginning to emerge and over the next two to three years the number of choices that are available to the end-user will continue to grow.

These new choices for the last mile include such services as xDSL, Cable Modem, Ethernet, and wireless. Additional choices will soon be available from power companies, low-earth orbit satellite systems and potentially even high-altitude aircraft based systems. Depending on availability, some locations may have fiber connected directly to their building that can provide enterprises with even more choices such as direct connectivity to a SONET service, Gigabit Ethernet service or even simply to lease dark fiber to build their own point-to-point networks.

WAN Core

Options within the WAN core have traditionally consisted of leased line, TDM, frame relay, or more recently, ATM. With the rise of the Internet a new choice has emerged, public IP network services.

Public IP services are typically being offered by Internet Service Providers (ISP’s) that are looking to leverage their existing networks to offer additional services. Public IP services either replace or supplement traditional frame relay, ATM and leased line services by tunneling private traffic within public shared networks based on IP. The largest drawback to public IP services has been the lack of guarantees of performance, however this is beginning to change as carriers improve their networks. Recently carriers such as UUNET have begun to offer service level guarantees for public IP networks that rival those offered by leased line and frame relay networks. As more providers improve their networks, public IP offerings should grow.

One of the keys to the future of the WAN is the development and deployment of multi-protocol label switching (MPLS). MPLS uses tags to mark specific IP traffic flows in order to provide varying classes of service. Using MPLS, carriers can engineer their networks to offer guaranteed service levels for multiple types of traffic flowing through their networks.

More recently there has been interest among carriers and other network service providers in using MPLS for IP-based VPN (Virtual Private Network) services. Carriers may use MPLS to "nail up" the equivalent of a permanent virtual circuit across an IP routed network; separate MPLS connections may even be established for each class of network service offered by the carrier. From the carrier's perspective, MPLS-based VPN connections (such as AT&T's "IP-Enabled Frame Relay") may be promoted as a replacement for frame relay PVC services. MPLS-based VPN's can also allow service providers to tunnel privately addressed networks through their own IP network. This removes one of the great barriers to VPN adoption by many large enterprises that have standardized on a private IP addressing scheme.

While only carriers and enterprises that own their own WAN infrastructure will likely deploy MPLS, MPLS will allow carriers to offer new services that will widen the choices available to their enterprise customers.

Convergence and Quality of Service

Perhaps no subject has received more press coverage in the last two years than voice, video and data integration. As the networking world has standardized on the IP protocol, and as new technologies have been introduced that improve network performance, it is only logical that organizations would look to consolidate their multiple networks for voice, video and data into a single network infrastructure based on IP. The architects of such networks believe that voice and video are simply data types that can be delivered over the same packet networks used to transmit other types of data.

However, the roadblock to convergence continues to be performance. Traditional IP data networks are based on best effort guarantees where no guarantee of reliable transmission is made. This type of network is fine for applications such as e-mail and web browsing but is insufficient to support application such as voice and video where delivery of data within a specified time range is required. In order to support applications such as video and voice, data networks must be converted to enable prioritization of voice and video data such that acceptable delays and jitter are achieved.

Fortunately, many efforts to deliver guaranteed Quality of Service (QoS) are underway. The two most important technologies under development are Diff-Serv and the previously mentioned MPLS.

Diff-Serv provides a way of using the Type of Service bit (TOS) in the IP packet header to mark a packet for a certain level of priority. Using Diff-Serv, a network can give higher priority to packets marked as voice or video, and lower priority to packets marked as web browsing or e-mail.

Both MPLS and Diff-Serv will likely be managed via policy-based networking. In early policy management implementations, network managers are able to centrally configure security and Quality of Service policy for the entire network from a single console. However, these policies are static and are not capable of changing based on certain network criteria such as load, user location, or day of the week.

However, in future policy management implementations, network managers will configure policy information for users and/or applications that is then stored in a directory (usually based on LDAP). Applications will request access to devices or certain quality of service levels (usually via RSVP (Resource Reservation Protocol). Network devices such as routers, switches (and in some implementations, network interface cards) forward the request to the policy manager using the COPS (Common Open Policy Services) protocol . Windows 2000 already supports APIs for using RSVP to request quality of service levels. The policy manager either grants or denies the request based on information stored within the policy management directory.

It should also be noted that the policy server model will also extend to security. In this case, the request/response protocol might be RADIUS for network authentication, authorization, and accounting functions. Policy servers will thus need to support multiple request/response protocols, including COPS and RADIUS.

The Desktop Management Taskforce (DTMF) is leading an effort to standardize how information about network objects is stored within the directory. The Directory-Enabled Networking (DEN) standard will attempt to create a common enterprise directory that can be shared among a variety of applications including messaging systems, network management systems, and network policy systems.

Technologies such as Diff-Serv and MPLS, along with initiatives such as DEN will eventually allow today’s data networks to deliver the guaranteed performance required for convergence of voice, video and data into the same network infrastructure. This will in turn lower management costs and reduce network complexity.

Conclusion

Within the scope of network architecture technology is constantly changing. While the pace of change of technology within the LAN has slowed in recent months, the pace of change in the WAN and in voice, video and data convergence is exploding. In order to maintain a competitive advantage, organizations must manage their network architecture to embrace, manage and exploit these changes.