Guest Column

Hardening MPLS Networks
(continued)

From a software perspective, a number of the edge router vendors have adopted a modular approach as well.  Separating software functionality into multiple, independent processes or threads helps to maximize efficiency and scalability, protect against system corruption and eliminate a single point of failure. A modular software design ensures no single process can consume all CPU resources, thereby preventing a single process failure from initiating a chain of events that could dramatically impact system performance.  This is critical to ensure stable and reliable operation under extreme network conditions.   

Next, let’s turn to the standards-based and proprietary techniques being used by vendors to improve reliability of routed and switched traffic carried across MPLS networks.   Network resiliency techniques are critical to ensure continued network performance during events where device-level resiliency is not sufficient.  These include:  an edge-core uplink failure, a core router failure and core line failures.  

BGP Graceful Restart:  Graceful restart is supported by most major IP/MPLS router vendors and has been defined by the IETF.  Graceful restart provides uninterrupted forwarding of packets when network stress forces a restart of the routing protocols by using the information stored in the device routing tables.  When service is restored, graceful restart notifies adjacent routers which then send the information required to build an updated routing table.  The benefits are that it enables hitless software upgrades and eliminates router flapping and subsequent packet loss.   In contrast, routers without graceful restart will allow peers to detect if a session goes down and must be restarted. The result is route re-computing and network-wide routing updates (also know as router flapping) which could result in packet loss.  The only downside to graceful restart is that adjacent (core and customer-based routers) must support this as well.  However, this has been a minimal issue since nearly all router vendors support this technique.

Non-Stop Routing:  This software technique is another approach to ensure IP packets continue to forward even if a route processor fails.  This is done by maintaining state information on a standby routing processor.  Since this technique does not require special communication with other routers, there is no need for standards work nor is there a need for adjacent routers to perform any software updates to successfully communicate changes (as with graceful restart).   Though nonstop routing shows promise, it is not widely supported by the vendor community.  In addition, there are questions about the scalability and reliability of this technique due to its requirement to copy all routing information on a standby processor. 

MPLS Fast Reroute:  MPLS Fast Reroute is designed to provide either local or global protection from link and node failures in an MPLS network without depending on routing protocols to reroute traffic.  The result is rapid restoration in as little as 50 milliseconds.  MPLS Fast Reroute works by creating backup paths ahead of time and then immediately switching to the backup path when a failure is detected.  The local protection techniques are most common on core routers and create a backup path for each link or router in the network.  Global protection techniques create a backup for each MPLS Label Switched Path (LSP) in the network and are commonly used on edge routers.

Proprietary Techniques:  Edge vendors, in particular, are developing proprietary techniques to enable diversion of ATM, Frame Relay, IP and MPLS IP VPN traffic around points of failure along an MPLS path in tens of milliseconds, which is equal to or faster than the under 50 millisecond recovery of SONET networks.   This is especially challenging at the edge of carrier networks, where path computation occurs, making restoration more challenging to implement.  This is enabled by pre-defined alternate paths that allow traffic to switch over to another MPLS path in the event of a link failure.  Other factors required to achieve this improved resiliency include hardware programming techniques that enable fast failover and the use of SONET timers.  

Some or many of the techniques described above will continue to evolve to improve the reliability of MPLS networks. MPLS has made significant strides to date, moving from a core traffic engineering technique to a way to enable service providers to turn their IP networks into a multi-service infrastructure capable of supporting all switched and routed data services.  In order for MPLS to reach its true potential, carriers and vendors must continue to work together to further raise the bar – making MPLS networks the reliable service provider infrastructure of choice. 

Stephen Vogelsang is Co-Founder and Vice President of Marketing for Laurel Networks.  Previously,  Vogelsang served as senior director of strategic and technical marketing at FORE Systems. He can be reached at sjv@laurelnetworks.com

<< Previous page