Timeline - October 2003: After a relatively flat year in 2003, sales in the core router market have started to pick up, as carriers begin deploying next-generation core IP routers that guarantee the magical “5 9’s” (99.999 per cent network uptime) offered by voice networks, bringing IP closer to becoming a multi-service infrastructure.

Still, IP reliability continues to be a highly contested and hotly debated issue among core router vendors and their customers. If you ask five different vendors to define IP reliability, you will probably hear five different answers. This is a proposal for a compelling and controversial presentation featuring different views on what exactly constitutes IP reliability and resiliency in today’s telco environment.

Non-stop forwarding/Protocol Extensions: Non-stop forwarding uses extensions to BGP (Graceful Restart), OSPF, IS-IS, LDP, and MPLS used to notify peer routers to continue forwarding and receiving packets, even though the route processor isn’t working. This method, promoted by the incumbent vendors, has both pros and cons. While it doesn’t require a back-up processor to store information, and there is no need to reboot the entire router, non-stop forwarding has been criticized for using stale information during failure, which can cause routing loops and black holes, and it also requires that surrounding routers adhere to separate extension standards for each protocol. Every router vendor must support protocol extensions for interoperability.

Non-stop routing/Stateful Failover/Fault-tolerant back-up: Non-stop routing uses backup route controllers, with a mirrored protocol exchange approach, to maintain all pertinent state information and maintain adjacencies with surrounding routers. The benefits include: preserved connectivity to peers, in-service software upgrades, no protocol extensions needed, and interoperability is not an issue. One con is that backup route processors will replicate any software error that occurred on the primary.

Software modularity: separating individual software processing flows and databases so that the failure of one process does not affect the other. This process can be taken off-line, upgraded or completely reloaded and restarted without impacting BGP, OSPF, ISIS or MPLS. Next-gen software architecture must function in a fully distributed manner across multiple controller cards, enabling scalable routing software, scalable control plane, and reliability, as no single controller failure will disrupt traffic flow throughout the system.

In-service software upgrades: the ability to upgrade a router without losing packets or service. This minimizes downtime for scheduled software upgrades.

Redundant hardware: All hardware components for next-generation routers must be fully redundant within a single system, including a multi-plane, load balanced switch fabric architecture that ensure line-rate traffic forwarding is maintained in the event of a switch card failure.

MPLS Fast Reroute: This is an IETF draft standard that provides reliability for IP/MPLS networks. Reroutes traffic around an outage within 50 to 100 milliseconds. Gives carriers flexibility to protect specific LSPs corresponding to premium services.

About Hyperchip: Hyperchip is a next-gen router vendor currently in field trials.

Louis is brings a wealth of relevant network experience to the Hyperchip team. Formerly a Distinguished Technical Member at UUNET Technologies/WorldCom, Louis is Hyperchip’s first Fellow, a title reserved to recognize individuals whose technical or scientific achievements have brought about meaningful advances in their fields. At UUNET Technologies/WorldCom, Louis was responsible for defining the basis of the DSL product implementation architecture for both residential and commercial products. The PPP-over-Ethernet (PPPoE) protocol - instigated and co-authored by Louis - became the de-facto standard for most residential DSL deployments in the world.

Louis's vast practical networking experience includes the network architecture, design and implementation of the UUNET DS3 backbone network in 1994 - the first large-scale deployment of a traffic-engineered Internet backbone. He also developed one of the first commercial IP-Multicast ISP-products while at UUNET.

Louis wrote his first IP stack for a UNIVAC 1108 mainframe in 1981 while on the staff at the University of Maryland, College Park. Starting in 1987. Louis was one in a group of 35 attendees at one of the first IETF meetings, and was the first chair of the Domain Name System working group within the IETF during 1987 and 1988.

Louis' vast experience has led him to numerous technical advisory board positions, including Allegro Networks, Procket Networks, Metricom, Quantum Bridge Communications, Tiara Networks, Appian Communications, and NetCore Networks. Louis earned a B.S. in Computer Science from the University of Maryland, College Park in 1981. He is a licensed amateur radio operator since 1974, a PADI certified scuba diver, and a serious amateur astronomer and astro-photographer.