Mainly due to the success of the Internet with its steady growth and ubiquitous reach, IP has for a long time shown its predominance as the preferred protocol and has more lately become the one protocol on which to base each application. However, network operators have been for many years compelled to deploy several other technologies to implement the services they wanted to sell. This led to the “networking onion” shown in the illustration here where various technologies are deployed in layers to finally support the provision of IP connectivity, as shown by the outer layer.
A wavelength division multiplexing (WDM) core is used to (i) maximize the transport capability of each fiber by transmitting over multiple optical channels on the same fiber and (ii) create long haul optical channels across multiple fibers among devices that are not directly connected. This is achieved through the deployment of (i) optical multiplexers and (ii) optical switches. Although more recently optical channels are being used to connect IP routers directly (as represented in the illustration here by the pipes extending from the WDM core to the IP outer layer) or even customer equipment, most often a circuit - switched network is built on them by means of SDH/SONET multiplexers and switches in order to enable the provision of lower - capacity channels. While the circuit - switching network can be used to sell services — specifically telephony and leased lines — directly to the customers, it does not feature the flexibility required for handling what is traditionally called data traffic. ATM (Asynchronous Transfer Mode) was standardized by ITU–T with the specific goal of realizing a Broadband Integrated Services Digital Network (ISDN) — that is, an infrastructure capable of providing the deterministic quality of service required by traditional telecom applications (such as telephony and videoconferencing) while efficiently supporting data traffic. Packet switching rather than circuit switching was chosen to achieve this goal, as highlighted by the term “asynchronous” that is juxtaposed to the synchronous nature of circuit switching. ATM failed to achieve the anticipated success and widespread reach because end users preferred the simplicity and low cost of technologies such as Ethernet and IP to the more sophisticated services of the more costly and complicated ATM. As a consequence, ATM has been confined to the operators ’ networks with few customers requiring ATM services (as represented by the pipes extending through the outer layers in illustration above ), possibly to interconnect their IP routers, or taking advantage of appealing commercial offers for circuit emulation over the ATM network to interconnect their PBXs (private branch exchanges).
Instead, operators resorted to their ATM infrastructure to interconnect their IP routers and provide IP connectivity services. Although the deployment of an ATM backbone to interconnect IP routers provides some advantages in controlling the traffic and hence the service provided, it is problematic due to the high cost of ATM interfaces on IP routers, which is due to the high complexity of ATM. Consequently, equipment manufacturers designed access devices with Frame Relay interfaces mainly used for the interconnection of IP routers (both the operators ’ and the customers ’ ones). Frame Relay is a convenient solution for the interconnection of IP routers because, being originally proposed by a consortium of IP router manufactures named the Frame Relay Forum, interfaces are available on virtually all high end routers at basically no cost. However, this required operators to add yet another layer to their “networking onion.” Such a multitude of technologies represents a significant cost for operators that need personnel trained on each of them, gear from several vendors, and spare parts for a large number of devices. Especially as the popularity and ubiquity of IP, further reinforced by the increasing deployment of IP telephony, results in most of the customers requiring IP connectivity services, operators would certainly benefit from getting rid of the “networking onion” in favor of an all IP network — that is, a mesh of IP routers interconnected by optical fibers. However, controlling the quality of the services offered on such a mesh interconnection of IP routers was until recently very complex, unless the network was engineered so that actual traffic would be a small fraction of the capacity. The sophisticated traffic management support offered by ATM enabled the realization of services with known quality through the ATM backbone while efficiently utilizing its resources. Consequently, commercial IP services have been provided by interconnecting peripheral IP routers through a partial mesh of virtual connections across an ATM backbone.
Could MPLS provide the solutions to the challenges highlighted above?
