Conceptual Underpinnings of Routing and Switching Protocols

To fully understand the operation of routing and switching protocols, you need to first understand several related concepts. These include:

• The traditional definitions of the terms routing and switching

• The new definition of the term routing as used by FC switch vendors

• The potential effect of loops within a topology

• The types of broadcast traffic that can be generated

• The difference between distance vector protocols and link-state protocols

• The difference between Interior Gateway Protocols (IGPs) and Exterior Gateway Protocols (EGPs)

In traditional IP and Ethernet terminology, the term routing describes the process of forwarding Layer 3 packets, whereas the terms bridging and switching describe the process of forwarding Layer 2 frames. This chapter uses the term switching instead of bridging when discussing Layer 2 forwarding. In both cases, the forwarding process involves two basic steps: path determination and frame/packet forwarding. Path determination (sometimes called path selection) involves some type of table lookup to determine the correct egress port or the next hop address. Frame/packet forwarding is the process of actually moving a received frame or packet from the ingress port to a queue associated with the appropriate egress port. Buffer management and scheduling algorithms then ensure the frame or packet is serialized onto the wire. The basic difference between routing and switching is that routing uses Layer 3 addresses for path determination, whereas switching uses Layer 2 addresses.

FC is a switching technology, and FC addresses are Layer 2 constructs. Therefore, FC switches do not "route" frames according to the traditional definition of routing. That said, the ANSI T11 FC-SW series of specifications refers to switching functionality, based on the Fabric Shortest Path First (FSPF) protocol, as "routing." FSPF must use FC addresses for path determination, therefore FSPF is actually a switching protocol according to the traditional definition. Moreover, many FC switch vendors have recently begun to offer FC routing products. Although the architectures of such offerings are quite different, they all accomplish the same goal, which is to connect separate FC-SANs without merging them. Since all FC routing solutions must use FC addresses for path determination, all FC routing solutions are actually FC switching solutions according to the traditional definition. However, FSPF employs a link-state algorithm, which is traditionally associated with Layer 3 routing protocols. Additionally, all FC routing solutions provide functionality that is similar to inter-VLAN IP routing. So, these new definitions of the term routing in the context of FC-SANs are not so egregious.

When a source node injects a frame or packet into a network, the frame or packet consumes network resources until it is delivered to the destination node. This is normal and does not present a problem as long as network resources are available. Of course, the underlying assumption is that each frame or packet will exit the network at some point in time. When this assumption fails to hold true, network resources become fully consumed as new frames or packets enter the network. Eventually, no new frames or packets can enter the network. This scenario can result from routing loops that cause frames or packets to be forwarded perpetually. For this reason, many routed protocols (such as IP) include a Time To Live (TTL) field (or equivalent) in the header. In the case of IP, the source node sets the TTL value, and each router decrements the TTL by one as part of the routing process. When the value in the TTL field reaches 0, the IP packet is discarded. This mechanism enables complex topologies in which loops might exist. However, even with the TTL mechanism, loops can cause problems. As the number of end nodes connected to an IP network grows, so does the network itself. The TTL mechanism limits network growth to the number of hops allowed by the TTL field. If the TTL limit is increased to enable network growth, the lifetime of looping packets is also increased. So, the TTL mechanism cannot solve the problem of loops in very large IP networks. Therefore, routing protocols that support complex topologies must implement other loop-suppression mechanisms to enable scalability. Even in small networks, the absence of a TTL mechanism requires the routing or switching protocol to suppress loops.

Note

The TTL field in the IP header does not limit the time each packet can live. The lifetime of an IP packet is measured in hops rather than time.

Protocols that support hierarchical addressing can also support three types of broadcast traffic:

• Local

• Directed

• All-networks

A local broadcast is sent to all nodes on the local network. A local broadcast contains the local network address in the high-order portion of the destination address and the all-nodes designator in the low-order portion of the destination address. A local broadcast is not forwarded by routers.

A directed broadcast is sent to all nodes on a specific, remote network. A directed broadcast contains the remote network address in the high-order portion of the destination address and the all-nodes designator in the low-order portion of the destination address. A directed broadcast is forwarded by routers in the same manner as a unicast packet until the broadcast packet reaches the destination network.

An all-networks broadcast is sent to all nodes on all networks. An all-networks broadcast contains the all-networks designator in the high-order portion of the destination address and the all-nodes designator in the low-order portion of the destination address. An all-networks broadcast is forwarded by routers. Because Ethernet addressing is flat, Ethernet supports only local broadcasts. IP addressing is hierarchical, but IP does not permit all-networks broadcasts. Instead, an all-networks broadcast (sent to IP address 255.255.255.255) is treated as a local broadcast. FC addressing is hierarchical, but the high-order portion of an FC address identifies a domain (an FC switch) rather than a network. So, the all-networks broadcast format equates to an all-domains broadcast format. FC supports only the all-domains broadcast format (no local or directed broadcasts). An all-domains broadcast is sent to D_ID 0xFF FF FF and is subject to zoning constraints (see Chapter 12, "Storage Network Security").

Note

Some people consider broadcast and multicast to be variations of the same theme. In that context, a broadcast is a simplified version of a multicast.

Each routing protocol is generally considered to be either a distance vector protocol (such as Routing Information Protocol [RIP]) or a link-state protocol (such as Open Shortest Path First [OSPF]). With a distance vector protocol, each router advertises its routing table to its neighbor routers. Initially, the only entries in a router's routing table are the networks to which the router is directly connected. Upon receipt of a distance vector advertisement, each receiving router updates its own routing table and then propagates its routing table to its neighbor routers. Thus, each router determines the best path to a remote network based on information received from neighbor routers. This is sometimes called routing by rumor because each router must make forwarding decisions based on unverified information.

By contrast, a router using a link-state protocol sends information about only its own interfaces. Such an advertisement is called a Link State Advertisement (LSA). Upon receipt of an LSA, each receiving router copies the information into a link-state database and then forwards the unmodified LSA to its neighbor routers. This process is called flooding. Thus, each router makes forwarding decisions based on information that is known to be accurate because the information is received from the actual source router.

In short, distance vector protocols advertise the entire routing table to adjacent routers only, whereas link-state protocols advertise only directly connected networks to all other routers. Distance vector protocols have the benefit of comparatively low processing overhead on routers, but advertisements can be comparatively large. Link-state protocols have the benefit of comparatively small advertisements, but processing overhead on routers can be comparatively high. Some routing protocols incorporate aspects of both distance vector and link state protocols. Such protocols are called hybrid routing protocols. Other variations also exist.

Routing protocols also are categorized as interior or exterior. An interior protocol is called an Interior Gateway Protocol (IGP), and an exterior protocol is called an Exterior Gateway Protocol (EGP). IGPs facilitate communication within a single administrative domain, and EGPs facilitate communication between administrative domains. An administrative domain can take the form of a corporation, an Internet Service Provider (ISP), a division within a government, and so on. Each administrative domain is called an Autonomous System (AS). Routing between Autonomous Systems is called inter-AS routing. To facilitate inter-AS routing on the Internet, IANA assigns a globally unique AS number to each AS.