Reliable Services in MPLS

ince the Internet was opened to commercial traffic in 1992, it has grown rapidly from an experimental research network to an extensive public data network. Demand is pushing the capabilities of today's Internet in several dimensions: transmission bandwidth, number of hosts, geographic size, and traffic volume. At the same time, the Internet is evolving from best-effort service toward an integrated or differentiated services framework with quality of service (QoS) assurances which will be necessary for many new applications such as voice over IP, videoconferencing, and multimedia. In recent years the industry has been searching for an approach to combine the best features of IP and asynchronous transfer mode (ATM), for example, IP routing with the performance and throughput of ATM switching. The Internet Engineering Task Force's (IETF's) classical-IP-over-ATM model treated IP as an overlay above ATM and defined logical IP subnets (LISs) over an ATM network [1]. This simple overlay approach allowed IP and ATM to work together without changes to either protocol, but did not take advantage of the strengths of ATM. Also, the approach was difficult to scale to many routers and was inefficient in certain aspects. The ATM Forum pursued an overlay approach with LAN emulation (LANE) and later multiprotocol over ATM (MPOA). The approaches used servers for address mapping and routing, and did not take advantage of QoS capabilities in ATM. The recent multiprotocol label switching (MPLS) approach is a convergence of various implementations of " IP switching " that use ATM-like label swapping to speed up IP packet forwarding without changes to existing IP routing protocols [2, 3]. Toshiba's Cell Switch Router (CSR) proposal in 1994 was perhaps the first proposal for an ATM switch that could be controlled by IP protocols rather than ATM signaling protocols. A CSR appears to be an IP router, but can select a flow for cut-through switching at the ATM layer to the next CSR. Ipsilon's proprietary IP Switch was essentially an ATM switch fabric controlled by an external switch controller running IP protocols. This was an example of the d a t ad r i v e n a p p r o a c h where persistent flows are automatically redirected through the ATM fabric. Cisco Systems' Tag Switching, an example of a control-driven approach, added a few innovations such as forwarding equivalence classes (FECs), a tag distribution protocol , and stacked tags. IBM's Aggregate Route-Based IP …