In my previous network design post we discussed modular network design. At its foundation modular network design is a system for creating repeatable network modules that can be deployed in a predictable manner as design metrics increase. These metrics could be number of subscribers, bandwidth consumption, number of services, and so on and so forth. In this article we are going to take a deeper look at Unified MPLS, a major enabler of modular network design.
What is Unified MPLS?
Unified MPLS, or end-to-end MPLS, is simple in concept. Layer 3 routing and MPLS are extended from the core all the way to the access edge (or even to the customer CPE). Subscriber connections, physical or virtually, are placed in various MPLS constructs: L3VPN, Pseudowire, or a VPLS domain. These constructs are used to connect the subscriber networks to the corresponding service modules off of the core network.
Why Unified MPLS?
So, great, we understand what it is. Now, why would we spend the money and resources to do this? It all comes down to simplification of service turn-up. Although a unified MPLS system may take a slightly longer time to initially deploy, making changes later and turning up services all become simpler and less time consuming after the day 1 deployment. For example, in a layer 2 access ring, to turn up a new VLAN for a customer (or perhaps for a new OLT) requires provisioning this additional VLAN on all devices on the ring. If a config mistake is made, such as not including the VLAN on a trunk port, the error may go undetected until a ring failure. This obviously would result in that entire VLAN being out of service during that cut. Also inserting a new node into the ring requires that all VLAN information be provisioned on the new node and the trunk ports.
On the flip side, with Unified MPLS, the access ring would be layer 3 and using a routing protocol such as OSPF. LDP would also be enabled to exchange MPLS labels. New customer VLANs would be terminated at a port (or a VLAN trunk port) connected to one of these layer 3 edge devices. Pseudowires would be used to “tunnel” that customers traffic back to a BRAS module or potentially directly to another layer 3 edge device (for a private LAN service). With this concept only the two end points for the circuit will require configuration and the configuration will be minimal – single line xconnect commands.
Inserting a new node into a layer 3 ring only requires configuring IP addresses on the physical interfaces, enabling OSPF for those two interfaces, and turning on MPLS. There is no need to have knowledge of the various pseudowires or services traversing the ring to provision a new node.
The customer connection or VLAN could also be placed into a VPLS domain for a multipoint scenario or into a layer 3 VPN/VRF for specialized routing requirements such as DIA. Each of the 3 MPLS services (PW, VPLS, and L3VPNs) can be placed into a MPLS-TE tunnel for fast reroute requirements or for customized paths through the network. High levels of QoS can also be utilized using the EXP bits in the MPLS labels. Also of these technologies can be automated using scripting and templates.
In the end Unified MPLS (and in the coming years, perhaps SDN) provides a mechanism to really accomplish a modular network architecture. It allows services to be layered/virtualized on top of the network infrastructure. Using VPLS and L3VPNs allow virtualized architectures to be created while pseudowires provide the plumbing for simpler services. Every customer connection can be provisioned in a templated way, consistent with all other services of that type in the network.