Network Analysis, Architecture and Design, Second Edition (The Morgan Kaufmann Series in Networking)
11.6 Routing
Routing is forwarding network-layer (IP) packets between IP end points across one or more networks. This is based on routers learning about the connectivity within and between networks and applying this connectivity information to forward IP packets toward their destinations. Routing is at a higher layer in the networking hierarchy and thus can have a much larger addressing scope. This is evidenced by the largest IP network, the Internet, which has worldwide scope and interconnects millions of devices.
Routing can be applied to the design as the sole interconnection mechanism or as an addition to a switching mechanism. In either case, it provides support for complex and large network designs (consider the size and complexity of the Internet) and offers features beyond the switching mechanism. From a scalability perspective, the same capacity-planning guidelines can apply to routing. Thus, when the maximum or sustained data rates of traffic flows are high enough to warrant segmenting a shared-medium network, routing could be used as the interconnection mechanism and routers can be placed between the networks.
Routers are also convenient devices to place security and management services, user interfaces, and accounting and administration. Routers support these services well, although their performance can suffer when too many services are placed on them. They are best placed at strategic locations in the network, such as where traffic flows are consolidated or where hierarchy occurs within the design. Since they route at the network layer, routers have access to network-layer information (e.g., source/destination IP addresses, port numbers, protocol information) that happens to be useful to support services. Since routers examine network-layer information, it is logical to use such information for other purposes, such as filtering traffic based on the source IP address or keeping counters of packet information for access by the Simple Network Management Protocol (SNMP).
When routers have support services in place, they make good locations to demark (generate and/or terminate) administrative boundaries. This is enhanced by routing protocols, which understand administrative boundaries through autonomous system (AS) numbers and routing areas. One feature of having such administrative boundaries is the ability to apply routing policies, higher-level abstractions of routing, to the network. As administrative features evolve, the ability to provide accounting information about the network will become increasingly important as a feedback loop in service support.
Another key aspect of routing is the inherent support for redundancy. By its nature, routing is closely tied to redundancy. In fact, without redundancy there really is no need for routing. As the degree of redundancy and/or the number of paths increases, routing increases in importance.
All of the aforementioned features help make routing a logical choice for external connectivity. When connecting to the Internet, IP routing is required. As trust declines in allowing outside connections to internal enterprise or personal resources, the security, monitoring, and administrative abilities associated with routers become essential in operating a network with external access. The combination of routing, external connectivity, and support services has led from placing routers between internal and external networks to the concept of the isolation LAN, also known as a demilitarized zone, as a boundary for external access. This concept, enhanced by the proliferation of Web information access and transfer, has led to a new term for such networks, the intranet. In the intranet in Figure 11.11, not only is there a separation of administrative boundaries, but information itself is also separated into external and internal resources.
Категории