Hardening MPLS Networks

By Steve Vogelsang
Vice President and Co-Founder
Laurel Networks, Inc.
02-September-2003

High availability is a must for all carrier networks, and MPLS-based networks are no exception.  MPLS first found its place in the core of ISP networks as a tool for traffic engineering and to add network resiliency to IP networks.   Now, MPLS has gained popularity among service providers due to its ability to enable a wide range of switched and routed data services from a single infrastructure, enabling carriers to collapse disparate networks over time.  The potential to enable new services while reducing cost and complexity has led providers across the globe to evolve to MPLS-based networks.  

According to a recent Infonetics report, the percentage of service provider respondents indicating MPLS deployment in some part of their network jumped from 47% in 2002 to 79% in 2003.   Yet, to be fully embraced by service providers as the basis for a single packet infrastructure, MPLS-networks must be as "hard" or robust as today’s ATM or Frame Relay networks.  That means they must far exceed the reliability of most IP networks (designed only to carry best-effort Internet traffic).  In particular, they must be reliable enough to meet the SLAs offered on current ATM and Frame Relay services or providers won’t migrate that traffic.  At the same time, MPLS networks must deliver extremely high uptime for new business services such as IP VPNs and reliably deliver real-time voice and video traffic to facilitate complete business communication services.   For business customers in particular, reliable network connectivity is more critical than ever as the business becomes more dependent on IT and network-based applications to communicate with remote offices, customers and suppliers.   

The need for MPLS reliability is most critical at the edge of service provider networks, which provide the link between customer networks and the provider.  In the core, MPLS-enabled routers often are deployed in pairs to create a redundant solution.  This redundancy provides built-in availability since the edge routers and other network elements can route around any core failure.  In contrast, at the edge a single router might directly interface to hundreds or thousands of customers, and any failure at the service provider edge can result in significant service disruption.  These network failures can translate into lost revenue since carriers often enter into service level agreements that specify guaranteed network uptime.

The past year has seen all vendors with MPLS edge offerings adding availability features to assure carriers that MPLS-based networks provide the stability required to carry all carrier data services. High availability features minimize the effect of network outages through device stability, redundancy, distributed hardware, software modularity and advanced network recovery capabilities.   Below is an overview of the major techniques used to improve network availability. 

Hardware redundancy is a clear place to begin and almost all vendors now feature redundant and hot swappable common components, switching fabrics, fans, power, etc.  This means that in the event of a component failure, the redundant equipment immediately takes over.   In fact, with most all vendors featuring redundant hardware, the focus of vendor development and standards work has turned to the software features required to maximize the availability of MPLS-based networks. 

System stability is another key component to building reliable MPLS networks.  For a system to remain stable in the face of network instability, it must contain significant processing power, since any routed MPLS network is constantly exchanging processor-intensive information.  A key to ensuring the required processing power is a distributed design.  This serves to separate routing from forwarding and other functionality, reducing the amount of state maintenance between the routing controllers since hardware state is maintained elsewhere.  It also serves to improve availability by eliminating a single point of failure. 

Next page >>

Page 1 of 2