Scientific learning and progress increasingly depend on high-speed academic networks that transfer huge amounts of data between geographically distant sites to facilitate collaboration. With more types of research using high-speed networks, requirements have become more diverse. For cutting-edge applications, academic networks need to enhance performance beyond current best efforts and provide a superior communication environment with sufficient bandwidth, small delays, and minimum delay variance. Collaborative research on shared networks calls for a closed user group environment to ensure security.
Academic networks tend to favor end-to-end circuits, i.e., dedicated lines, to create high-performance environments.1,2 These networks use lower-layer devices for circuit services in addition to existing Internet protocol (IP) routers for packet (IP and Ethernet) services. They need separate wavelengths for circuit services, and because it is difficult to predict precise demands, it might be considered a risky investment to develop this architecture for such services nationwide. Therefore, a more highly integrated architecture, devoted to both circuit and packet services, is highly desirable.
Figure 1. The new academic network deploys 75 edge nodes, 12 core nodes, and 40Gbps lines. L1: Layer 1. IP: Internet protocol.
My colleagues and I started full-scale operation of a wholly integrated network for the Japanese research and education community, SINET3, in June 2007.3 SINET3's architecture accommodates all of the available services in a single network and flexibly manages resource assignment depending on service demands. SINET3 attains enormous bandwidth by using the world's fastest lines, at 40Gbps, and provides an unparalleled variety of network capabilities, such as multiple layers (layer 3: IP, layer 2: Ethernet, and layer 1), an enriched virtual private network (VPN), enhanced quality of service, and bandwidth on demand (BoD).
SINET3 is an advanced optical and IP hybrid network composed of 75 layer-1 switches, i.e., next-generation synchronous digital hierarchy (SDH) devices, and 12 IP routers (see Figure 1). Circuit services are accommodated in the same SDH lines as packet services, and the assigned bandwidth for each can be changed hitlessly (with no negative effect for users) using sophisticated SDH technologies. The network forms multiple loops in the backbone that enable fast recovery functions.
SINET3 incorporates the latest advances to cost-effectively achieve the extensive menu of services offered (see Figure 2). It employs SDH methods such as generic framing procedure, virtual concatenation (VCAT), and link capacity adjustment scheme (LCAS) suitable for multilayer services and flexible, reliable assignment of resources. The network uses multiprotocol label switching (MPLS) and logical router technologies, which separate the VPN instances in terms of routing, signaling, and forwarding, to implement a variety of VPN services on the IP routers. Effective use of the failure detection capabilities of layer-1 switches and the fast detour functionalities of both the IP routers and layer-1 switches enables quick service recovery.
The network provides BoD as part of the layer-1 services, and gives users a protocol-free and completely exclusive communication environment. Individuals can specify the destinations, duration, and bandwidth of data transmission with virtual container-4 (about 150Mbps) granularity. These services rely on an intelligent layer-1 BoD server through which the users can directly request layer-1 path setups between generalized MPLS (GMPLS)-based layer-1 switches. The BoD server can receive users’ requests, schedule accepted reservations, calculate optimal routes based on constraints such as the minimum end-to-end delay, and trigger layer-1 path setup and release.
Figure 2. Numerous advanced networking technologies, such as GFP/VCAT/ LCAS, logical routers, MPLS, and GMPLS, have been combined to provide a variety of services. STM: Synchronous transport module. LR: Logical router. LCAS: Link capacity adjustment scheme. MPLS: Multiprotocol label switching. GMPLS: Generalized MPLS. VLAN: Virtual local area network. Mux: Multiplexer. GE: Gigabit Ethernet.
In summary, SINET3 features numerous advanced networking functions, such as GFP/VCAT/LCAS, logical router, MPLS, GMPLS, and fast detour capability, in addition to 40Gbps lines. We have just started BoD services and will report on the detailed evaluation and enhanced features of these dynamic circuit services in the near future.
National Institute of Informatics (NII)
Shigeo Urushidani is a professor and responsible for the design and development of SINET3. He joined NII in April 2006 after eight years of experience as a visiting professor there. Before joining NII, he worked for the laboratories of Nippon Telegraph and Telephone, where he engaged in research and development on network service systems for asynchronous transfer mode networks, advanced intelligent networks, high-speed IP/MPLS networks, and GMPLS-based optical networks.