It would be both incomplete and inaccurate to preview the internetworking world of 2002 without acknowledging the horrendous body count resulting from the recession of 2001. Certainly, previous downturns have led to the demise of specific products that perhaps deserved a chance in the market. But looking for new ideas in internetworking at the end of this year was an awful lot like seeking out the best and brightest in an area racked with carpet bombing.
So does this mean there is nothing new under a bleak and wintry sun? Hardly. The combination of research institutions involved in Internet2, a few hardy Internet Protocol startups like Packet Design Inc. and Allegro Networks Inc., and some surviving research efforts at Cisco and Unisphere have ensured that the 25-year-old beast known as IP routing has some new wrinkles.
As heightened interest in security has spurred the corporate use of virtual private networks, Allegro is pursuing a particularly interesting space by promoting clustered routers in a rack, implementing a concept the company calls "real private networks." Rather than creating artificial tunnels of IP flows through use of the IP Secure protocol, Allegro is creating private-network domains through multiple physical-router systems operating in parallel. By looking at a new hardware method of segmenting router domains, Allegro may have identified a method of scaling edge routers more economically than typical architectures from the likes of Cisco or Juniper.
New ideas are being generated for mixing dense wave-division-multiplexing (DWDM) equipment, which aggregates analog optical wavelength channels with both time-division-multiplexed Sonet and IP routing equipment. In early December, Cisco upgraded its IP+Optical program by pulling together elements of its Cerent Sonet and packet-provisioning system, its Qeyton metropolitan DWDM system and the Cisco homegrown routers into a new entity the company calls Complete Optical Multiservice Edge and Transport, or Comet. While critics charge that Cisco is merging disparate elements in a program largely centered on marketing, the company is coming up with legitimate strategies for handling optical and IP transport.
This heightens the importance of Cisco's recent decision to pour new investment into European research efforts on services for IP Version 6. It also underscores Cisco's decision to be a strong champion of generalized multiprotocol label switching. The original MPLS, a method of swapping IP header "labels" in order to manage IP flows, was invented at Cisco in the early 1990s. A spin of MPLS dedicated to optical-wavelength management pioneered by Daniel Awduche, then at UUNet/WorldCom, and dubbed multiprotocol lambda switching was rechristened G-MPLS by the Internet Engineering Task Force. The notion of using G-MPLS with hybrid switch architectures, as a way of controlling optical aggregation through an electronic control plane, will be one of the key areas to watch in 2002.
And lambda-switching author Awduche may be one of those leading the charge. Awduche left the carrier community to join a metropolitan optical-switching startup, Movaz Networks Inc., that may be one of the few fledgling companies to survive in what looks like a very tough 2002. Movaz is trying to push carriers toward mesh topologies in metropolitan networks, using G-MPLS as a control-plane tool for embedded signaling, routing and link management.
Ciena also is in a good position in metropolitan aggregation, particularly after the acquisition of switch developer LightEra, which gave Ciena its CoreDirector hybrid switch. Ciena is promoting the virtual concatenation of Sonet rings as a way to bring carriers into the packet-switching world without tossing away existing Sonet equipment. In that sense, Cisco's use of the Cerent architecture and Ciena's use of CoreDirector bear similarities.
Coming in from the Ethernet side of the street are companies such as Foundry Networks Inc. and Extreme Networks Inc., which have added hardware redundancy to their switches and scaled them up in capacity, to make them more appropriate for carrier customers. When telephony traditionalists claimed that Ethernet was not ready to handle full Sonet protection switching and lifeline telephone services, a cluster of startups, along with Cisco, formed the group that eventually became the IEEE-802.17 resilient packet ring effort.
At the edges of the IP cloud, closer to the end customers, the action in edge routing will be lively in 2002, as existing vendors like Cisco and Juniper implement MPLS systems. To push broadband service to end customers, a few carriers, such as SBC Communications, will push fiber further out into the network via passive optical networks. But it's not economical to extend the PONs all the way to the individual customer.
Carriers in the post-2001 meltdown will want to be able to add bandwidth in very cost-effective ways, and that means using quality-of-service (QoS) on a regular basis. Edge equipment will have to be able to dynamically provision bandwidth in chunks down to 64-kbit/second (DS-0) increments, all in near real-time.
This isn't just an issue for the traffic-management chip set people. The pioneers in IP routing software will be upgrading specialized protocols for QoS, flow management, security and multimedia multicasting. Keeping software in the edge router updated will be as important as making sure the line cards can handle QoS for MPLS and Differentiated Services.
The roster of internetworking players changes daily, often in unpredictable ways. And when capacity demands start to pick up among carriers in mid-2002, the directions those carriers head in equipment purchases will be equally unpredictable. Recovery in telecommunications will happen in sudden fits and starts, with certain technology contenders hot today and obsolete tomorrow. Observers will keep their eyes trained on the likely inflection points G-MPLS, parallel routing, RPR, soft switches. But sudden equipment ramps and shifts in favored architectures will take place faster than the best market analysts will be able to predict.