Why MPLS matters in carrier networks
Unless you've been pulling a Rip Van Winkle for the past 18 months, you've noticed that Multi-protocol Label Switching has become the technology du jour.
Virtually every major provider either has deployed it or has plans to do so. There are books and conferences devoted to the technology. So what's all the fuss about? From a user perspective, mighty little.
MPLS offers basically the same thing available previously from technologies such as ATM and frame relay: relatively reliable and secure transport across the WAN. However, to service providers, MPLS promises increased control, simplicity and manageability.
To understand what MPLS does and does not accomplish, you need to step back and revisit data networking architectural principles. One of the biggest challenges in building a network lies in ensuring traffic has enough bandwidth to get from point A to point B. In the old days before IP, network managers solved this problem by engineering the network to support anticipated traffic flows - a process called, naturally enough, traffic engineering. In this model, all the routing intelligence happened outside the switching fabric, in complex modeling software or engineering craniums. Engineers told the switches where the traffic should go, and the switches obeyed.
However, the paradigm shift introduced by IP is that the network (more specifically, the routing elements that forward the traffic) figures out the best path across the network. In essence, routers do traffic engineering.
This approach has many strengths but also major weaknesses. When networks get too large, discovering and selecting routes becomes slow, awkward and ineffective. For this reason, by the late 1990s, most large ISPs had created two-tier architectures, with an outer ring of intelligent routers communicating across a switched (typically ATM) core, implementing traffic engineering at the core.
There are two problems with this approach. First is the well known "cell tax," the bandwidth overhead resulting from segmenting large IP packets into 53-byte ATM cells. In addition, service providers must manage and administer multiple networks of devices (optical transmission, ATM switching and IP routing).
MPLS addresses both problems. In essence, it provides for ATM-like traffic engineering across an all-IP backbone, eliminating the cell tax and reducing the number of networks that must be managed - both net gains for service providers. Additionally, MPLS by design does not interfere with user address space. This means that, unlike in a "pure" routed IP network, users can maintain their addresses, even using private (net10) addresses. Again, this feature was and is available over a frame relay or ATM network; the difference in this case is that it's available across an IP one.
All that said, MPLS isn't a complete panacea. Ensuring the scalability and security of MPLS networks is something of a black art, requiring sophisticated network design. But overall, MPLS gives service providers increased control over their networks - and that can only translate to better service quality for users.
RELATED LINKS
Johnson is senior vice president and CTO for Greenwich Technology Partners, a network consulting and engineering firm. Her column appears biweekly. She can be reached atjohna@greenwichtech.com.
