5.3. Control Plane Aspects in IP Networks

Figure 5-10 illustrates a simple IP network. Here, the IP networking infrastructure consists of Local Area Networks (LANs; vertical and horizontal bars) interconnected by routers, whose task is to forward packets from a source end-system to a destination end-system. An end-system is typically a host attached to a LAN (routers can also be sources and destinations of IP packets, specifically those related to the control plane). An IP packet thus contains a source address and a destination address, among other information, as shown in Figure 5-11. Each router maintains a routing table (also called forwarding table) that lists the address of the next router (or end-system) corresponding to each known destination address. Thus, a router receiving an IP packet looks up its routing table, determines the next hop, and forwards the packet to that address. This way, a packet is forwarded hop by hop to its destination.

The above is a simple description of the transport plane of IP networks. The control plane that enables the forwarding of IP packets is more complex. Specifically, the control plane has the following components of interest to us:

Understanding these aspects of the IP network control plane is essential for developing an IP-centric control plane for optical networks.

5.3.1. Addressing

In IP networks, each network interface and/or node (router or host) must be assigned a unique IP address. The uniqueness is defined within the scope of the routing domain, for example, a private network or the entire Internet. Address assignment can be manual or automatic, for example, using the Dynamic Host Configuration Protocol (DHCP) [Alexander+97]. The former method is typically used to assign addresses to routers and their interfaces. The latter is used to assign addresses dynamically to hosts that may connect and disconnect from an IP network, for example, over a dial-up connection. Either way, the assignment of unique addresses allows a packet to be routed to the correct destination.

IP addresses, as shown in Figure 5-11, are 32 bits long. An IP address prefix is a string whose length is less than or equal to 32 bits. A prefix represents all IP addresses that have the same values in the corresponding bit positions. Figure 5-12 illustrates a 24-bit IP address prefix and the IP addresses represented by the prefix. Prefixes are used as shorthand to represent multiple IP addresses, especially in IP routing protocols. By using prefixes, the amount of information maintained and propagated by an IP routing protocol decreases thereby improving the scalability of routing. A prefix is typically represented as a 32-bit number and a mask that indicates which bits are “active” in the prefix. Figure 5-12 shows the representation of a prefix in this manner.

A given IP address could be included in more than one prefix. For example, considering Figure 5-12 again, the address 198.16.101.12 is included in the 24-bit prefix 198.16.101 shown. The same address is also included in the 16-bit prefix 198.16. When there are multiple prefixes including a given address, the longest matching prefix is the one with the most number of bits. In the current example, the 24-bit prefix is longer than the 16-bit prefix. An IP routing table may contain multiple prefixes that include a given destination address. In this case, the entry corresponding to the longest matching prefix is considered most specific. Thus, a router with multiple prefixes matching a given destination address will typically select the next hop corresponding to the longest matching prefix.

Finally, IP addresses can be classified into unicast or multicast. A multicast IP address is a logical address denoting more than one physical destination. An IP packet sent to a multicast address will be received by all the destination nodes. Support for multicast requires specialized routing procedures within the network [Tanenbaum02].

5.3.2. Discovery

A router must determine whether it has direct connectivity to other routers. This information is essential for building its routing table. The discovery of adjacent routers is typically a part of IP routing protocols. Under one realization, each router periodically sends a keep-alive message over each of its interfaces. This message contains the identity of the sender (e.g., its IP address), and it is addressed to a special “all routers” multicast IP address. A router receiving such a message becomes aware of the existence of the sending router in its neighborhood.

5.3.3. Routing

Each router in an IP network must know whether a given IP destination address is reachable, and if so, to which next hop the packet should be forwarded. This is essentially the information available in the routing table. The routing table could be built automatically using a distributing IP routing protocol, or by manual configuration. The details of IP routing protocols are described in Chapter 9. Here, we just note briefly the functions of an IP routing protocol and the different types of protocols.

An IP routing protocol allows a router to determine directly reachable IP addresses, propagate this information to other routers in the network, and use this information, along with information received from other routers, to build its own routing table. The global Internet consists of a large set of interconnected, administratively independent IP networks. IP routing is therefore structured into intradomain and interdomain routing. An intradomain routing protocol is used within an independently administered network, also called an Autonomous System (AS). For historical reasons, such a protocol is also referred to as an Interior Gateway Protocol (IGP). Examples of IGP are the Routing Information Protocol (RIP) [Hedrick88] and the Open Shortest-Path First (OSPF) protocol [Moy98]. An interdomain routing protocol is used to route packets across different ASs. Historically, such a protocol has been referred to as an Exterior Gateway Protocol (EGP). An example of an EGP is the Border Gateway Protocol (BGP) [Rekhter+95]. It should be noted that although both IGPs and EGPs are used for building routing tables, the criteria used in these protocols for selecting routes and the manner in which routing information is propagated are different. One aspect that is considered important in EGPs is the aggregation of IP addresses propagated across ASs. Specifically, address assignment within an AS must be such that it must be possible to summarize the addresses with a few prefixes. If this is not the case, the routing tables in routers running EGPs tend to grow large, affecting the scalability of the EGP. IGPs and EGPs are further described in Chapter 9.