Consolidation and virtualization are two important trends in both service provider and enterprise networks. These are really just two aspects of the same trend: The economization of network infrastructure to reduce both capital expenditures and operational cost.

Consolidation is the reduction of multiple, service-specific infrastructures (Internet and intranet, voice, video, dedicated point-to-point circuits, etc.) to a single multipurpose infrastructure. That aspect of consolidation has been around for a long time (X.25, Frame Relay, and ATM were early enablers), but the trend is continuing now by consolidating multiple nodes in the multipurpose infrastructure into fewer multiservice nodes. Virtualization is the logical approach by which the consolidated infrastructure supports multiple services and fulfills the unique characterizations of each.

There are a few essential requirements for virtualization:

1. Transparency: Services must be able to share network resources without any service negatively impacting another service.

2. Privacy: The information used by one service must not be visible to other services.

3.  Resiliency: The infrastructure must be able to bypass or quickly recover from failures, congestion, and performance slowdowns so that the impact on sensitive services is minimized.

4. Multiple logical topologies: The consolidated infrastructure must provision service demarcations wherever they are needed but only where they are needed.

5.  Optimized resource utilization: The available bandwidth, latencies, processing, and memory of the underlying infrastructure must be efficiently used, ensuring that a situation does not arise in which some parts of the network are overutilized while other parts are underutilized.

MPLS is the technology of choice for consolidation and virtualization in service provider networks. Its failure protection and recovery mechanisms, traffic engineering capabilities, flexible connectivity options, VPN features, and manageability give service providers the agility to serve a wide variety of customers and to quickly roll out new services.

MPLS seldom, however, reaches beyond the service providers’ PoPs. Although MPLS is a mature, proven technology, enterprises, for the most part, stick with Ethernet to achieve virtualization in their LAN and data center infrastructures: VLANs for privacy, trunking for logical topologies, and QinQ to overcome scaling limitations. Manageability and efficient resource utilization tend to be more challenging in enterprise networks than in service provider networks. Then, there is Spanning Tree, a necessary evil in redundant Ethernet topologies but almost universally hated for its slow recovery times, irascible behavior, and troubleshooting difficulty.

MPLS is still new to most enterprise network operators; as a result, most of their networks continue to be designed according to the conventional wisdom of the mid-1990s.

But MPLS is beginning to find its way into some large data centers, such as those of the New York Stock Exchange. Vendors are also beginning to recognize the potential of MPLS in enterprise networks; Juniper, for one, has been touting the strengths of its MPLS-capable MX as both a Metro-E and a data center product.  Cisco’s MPLS-capable Catalyst 6500 has long been a data center stalwart. Force10’s new ExaScale E-Series switches, built for the data center market, have MPLS capabilities in their ASICs (although not yet activated) in anticipation of a growing MPLS presence in the market.

It’s doubtful that you’ll see MPLS in smaller LANs anytime soon, if ever. But the technology will grow in larger enterprise networks as a superior solution to the complexities of virtualization, leaving Ethernet to be what it should be: A fast, inexpensive, uncomplicated data link technology.