Special Report: Next Generation Internet Applications
Improvements for IP routing and QoS currently hinge upon the IETF MPLS standard for label switching of IP packets. MPLS provides ATM-like features by providing high speed switching of packets along a defined path for each session. Further development is needed for the proper integration of MPLS and associated standards like Differentiated Services (Diff-Serv) into high capacity IP routers. Advances in MPLS may eventually spill over into the optical domain.
The photonic counterpart to MPLS, MP'S, may be used within all-optical networks. MP'S applies label techniques to wavelength switched optical paths. Each wavelength serves as its own label. Both MPLS and MP'S must perform control functions like addressing, routing, signaling, and survivability. Used in combination, MPLS and MP'S may address the missing elements in today's IP-over-DWDM implementations. However, implementation of MP'S will be quite different from MPLS. Control functions within MP'S are different than MPLS. The control and data planes must be separate in an all-optical network because of the very limited number of optical switching and signal processing devices currently available.
There are three architectures under consideration for controlling the optical layer: router-centric, client-server, and peer-to-peer.[29][30] See Figure 28 for an illustration of these architectures. Router-centric architecture places all of the switching intelligence in high-speed electronic routers and dedicates DWDM light paths between the router. High capacity routers assume all control of packet movement.
Switching within the optical layer is static with optical control functions relegated to long term path provisioning, protection, and restoration.
Router-centric architectures suffer from a lack of scalability as optical channel bandwidths double every nine months, making electronic routers inadequate to the task of managing huge optical bandwidths. Advances in electronics are not increasing IP routing and ATM switching capacity fast enough to keep pace with DWDM capacity growth. Therefore, a router-centric architecture is not a good long-term approach for high-speed core networks.
Client-server, also called "interdomain," architectures keep the electronic routing layer separate from the optical DWDM layer. Client-server networks have separate routing instances within the IP and the optical layers. Routing information is shared between the instances. Routers and optical cross connects (OXCs) have unique IP addresses and a separate control plane for exchanging network, link status, provisioning, and restoration information. There is signaling between the two layers.
Client-server architectures are considered to be the most viable architecture in the short term. Eventually, new more elaborate standards for interfacing IP and DWDM networks will develop, creating a more unified, peer-to-peer architecture. Peer-to-peer networks resemble the separate control and data planes in the SS7-based telephone network as illustrated in Figure 28. All electronic routing and optical DWDM layers are controlled from a single, unified control plane.
Previous Next Table
of Content for report: Next Generation Internet Applications Home
