Is MPLS Right for You?

Multiprotocol label switching (MPLS) is a relatively new way to deliver network services, embraced by both carriers and enterprises. When carriers sell point-to-point or multipoint network services, MPLS is just a buzzword. Users buy a provisioned network (usually ­delivered as Ethernet WAN service) and don't really care whether the carrier used MPLS to build it.

Any doubts that MPLS is popular with carriers can be settled by looking at the massive rise in end-user use of Ethernet metro area and wide area network (MAN/WAN) services and the corresponding deinstallation of end-user connections such as Frame Relay and ATM. Network managers strongly prefer to use Ethernet technology throughout their networks, whether LAN or WAN. MPLS is proving an ideal way to deliver that technology, and moves much of the difficult WAN management from end user to carrier. For the network manager, the capital costs are dramatically lower, and the consistent simplicity of Ethernet makes everything easier, from design to deployment to debugging.

MPLS itself is not just for carriers. The underlying technology also can prove useful to enterprise network managers, especially in campus and multibuilding environments within higher education.

Traditional IP networks use physical locations as convenient places for routing: Each floor, each building or each wiring closet becomes an IP subnet or set of IP subnets. Traffic then flows back toward central backbones and common data centers. Environments with intermixed end users all focused on a few data centers can continue to use this traditional topology and won't find much value in switching to MPLS.

There are some strong reasons to use MPLS in more complex campus environments. For network managers who need even one of the following features, MPLS is the right technology to deploy.

Workgroup Isolation — for implementing multiservice networks using VPNs:

In IP environments, the key factor pushing toward MPLS technology is the need for workgroup isolation and special workgroup-type protocols (such as multicast or experimental protocols). For example, if a workgroup is spread across buildings or floors but still needs to be networked as if on the same subnet, then MPLS-based virtual private networks are the ideal tool to extend Layer 3 subnets across Layer 2 networks.

MPLS-based VPNs also are very useful when workgroups have private data centers, or when a workgroup has specific security and firewall requirements. By linking workgroup members to a single virtual network, MPLS VPNs simplify firewall design and topology.

MPLS can deliver VPNs to workgroups that appear as large Layer 2 networks, usually called pseudowire, virtual leased line or virtual private LAN service; or (preferably) as independent and isolated Layer 3 networks — usually called L3VPN.

Layer 2 MPLS VPNs historically have been important for certain types of networking, such as linking highly virtualized data centers across buildings, although this requirement is falling away as newer protocols such as VXLAN move this particular interconnection from Layer 2 to Layer 3. Network managers who are considering MPLS VPNs should heed the age-old advice: "Route where you can, switch where you must," and deliver Layer 3 VPN services instead of Layer 2 VPNs wherever possible.

The Carrier/Subscriber Business Model — in which the telecommunications department functions as a pseudo-carrier:

Because MPLS is a technology designed for carriers, it is also very useful in large campus environments where the telecommunications department acts as a carrier to departmental subscribers.

MPLS technology allows telecommunications teams to deliver network services to the edge of a user or departmental network, which gives end users huge leeway in designing and managing their own departmental networks. Issues such as addressing and security within that network become nicely partitioned so that telecommunications groups operate in hands-off mode, delivering network services without getting into the details of how individual group networks are managed.

With MPLS-based networks, the customer edge/provider edge boundary is very clear and gives an obvious demarcation point between the telecom group and the departmental network. This helps to simplify debugging by clarifying responsibilities and making network services service-level agreement definitions possible. Even if the telecommunications department doesn't act as a carrier to end-user departments, network abstraction is a strong reason to adopt MPLS technology.

When buildings (or even wiring concentrators) are ­connected using MPLS, the entire underlying ­interbuilding MAN/WAN network can be swapped out, upgraded or replaced without the building network manager having to make any changes, except perhaps to move an Ethernet cable from one port to another.

Network abstraction as provided by MPLS allows the end-user part of the network to develop at a different pace from the interconnection part of the network.

Traffic Engineering and Fast Reroute — when powerful routing and congestion control are required:

While Ethernet and TCP/IP's "best effort" networking are very successful in small networks, large networks often require more sophisticated routing and quality-of-service management. MPLS traffic engineering does a better job of providing highly differentiated services than simple IP routing can.

MPLS isn't a traditional QoS enforcer, dropping selected packets when circuits get congested, but offers a type of traffic engineering that can reserve diverse paths through the network based on application requirements and available bandwidth. When multiple links are present (such as for reliability purposes), MPLS traffic engineering offers the ability to increase network use on campus networks, making better use of capital investments.

All of the MPLS features laid out here are more advanced, and many higher education networking teams can make use of MPLS without diving this deep. However, core traffic engineering features within MPLS, such as resource reservation, can also prove very useful as demands on the campus network increase.

For example, when different network users have wildly different requirements for their traffic, MPLS resource reservation and traffic engineering can ensure that the high-energy physics group that needs 40 gigabits per second of guaranteed bandwidth to an Internet2 connection doesn't conflict with a student services group that requires low-latency, low-jitter bandwidth for VoIP services.

In environments where small network outages have big consequences, faster network convergence is a good reason to add fast reroute to an MPLS backbone.