Overview of Basic Router Platform Architecture and Processes
To understand how switching works, it helps to first understand the basic router architecture and where various processes occur in the router.
Fast switching is enabled by default on all interfaces that support fast switching. If you have a situation where you need to disable fast switching and fall back to the process-switching path, understanding how various processes affect the router and where they occur will help you determine your alternatives. This understanding is especially helpful when you are troubleshooting traffic problems or need to process packets that require special handling. Some diagnostic or control resources are not compatible with fast switching or come at the expense of processing and switching efficiency. Understanding those resources can help you minimize their effect on network performance.
Figure 1-1 illustrates a possible internal configuration of a Cisco 7500 series router. In this configuration, the Cisco 7500 series router has an integrated Route/Switch Processor (RSP) and uses route caching to forward packets. The Cisco 7500 series router also uses Versatile Interface Processors (VIPs), a RISC-based interface processor that receives and caches routing information from the RSP. The VIP card uses the route cache to make switching decisions locally, which relieves the RSP of involvement and speeds overall throughput. This type of switching is called distributed switching. Multiple VIP cards can be installed in one router.
Figure 1-1. Basic Router Architecture
Cisco Routing and Switching Processes
The routing, or forwarding, function comprises two interrelated processes to move information in the network:
Making a routing decision by routing
Moving packets to the next-hop destination by switching
Cisco IOS platforms perform both routing and switching, and there are several types of each.
Routing
The routing process assesses the source and destination of traffic based on knowledge of network conditions. Routing functions identify the best path to use for moving the traffic to the destination from one or more of the router interfaces. The routing decision is based on a variation of criteria such as link speed, topological distance, and protocol. Each separate protocol maintains its own routing information.
Routing is more processing intensive and has a higher latency than switching as it determines path and next-hop considerations. The first packet routed requires a lookup in the routing table to determine the route. The route cache is populated after the first packet is routed by the route-table lookup. Subsequent traffic for the same destination is switched using the routing information stored in the route cache. Figure 1-2 illustrates the basic routing process.
Figure 1-2. The Routing Process
A router sends routing updates out to each of its interfaces that are configured for a particular protocol. It also receives routing updates from other attached routers. From these received updates and its knowledge of attached networks, it builds a map of the network topology.
Switching
Through the switching process, the router determines the next hop toward the destination address. Switching moves traffic from an input interface to one or more output interfaces. Switching is optimized and has a lower latency than routing because it can move packets, frames, or cells from buffer to buffer with a simpler determination of the source and destination of the traffic. It saves resources because it does not involve extra lookups. Figure 1-3 illustrates the basic switching process.
Figure 1-3. The Switching Process
In Figure 1-3, packets are received on the Fast Ethernet interface and destined for the FDDI interface. Based on information in the packet header and destination information stored in the routing table, the router determines the destination interface. It looks in the protocol's routing table to discover the destination interface that services the destination address of the packet.
The destination address is stored in tables, such as ARP tables for IP and AARP tables for AppleTalk. If there is no entry for the destination, the router will either drop the packet (and inform the user if the protocol provides that feature), or it must discover the destination address by some other address resolution process, such as through the ARP protocol. Layer 3 IP addressing information is mapped to the Layer 2 MAC address for the next hop. Figure 1-4 illustrates the mapping that occurs to determine the next hop.
Figure 1-4. Layer 3-to-Layer 2 Mapping
Basic Switching Paths
Process Switching
Fast Switching
Distributed Switching
NetFlow Switching
Process Switching
In process switching, the first packet is copied to the system buffer. The router looks up the Layer 3 network address in the routing table and initializes the fast-switch cache. The frame is rewritten with the destination address and sent to the exit interface that services that destination. Subsequent packets for that destination are sent by the same switching path. The route processor computes the cyclic redundancy check (CRC).
Fast Switching
When packets are fast switched, the first packet is copied to packet memory and the destination network or host is found in the fast-switching cache. The frame is rewritten and sent to the exit interface that services the destination. Subsequent packets for the same destination use the same switching path. The interface processor computes the CRC.
Distributed Switching
Switching becomes more efficient the closer to the interface the function occurs. In distributed switching, the switching process occurs on VIP and other interface cards that support switching. Figure 1-5 illustrates the distributed switching process on the Cisco 7500 series.
Figure 1-5. Distributed Switching on Cisco 7500 Series Routers
The VIP card installed in this router maintains a copy of the routing cache information needed to forward packets. Because the VIP card has the routing information it needs, it performs the switching locally, making the packet forwarding much faster. Router throughput is increased linearly based on the number of VIP cards installed in the router.
NetFlow Switching
NetFlow switching enables you to collect the data required for flexible and detailed accounting, billing, and chargeback for network and application resource utilization. Accounting data can be collected for both dedicated line and dial-access accounting. NetFlow switching over a foundation of VLAN technologies provides the benefits of switching and routing on the same platforms. NetFlow switching is supported over switched LAN or ATM backbones, allowing scalable inter-VLAN forwarding. NetFlow switching can be deployed at any location in the network as an extension to existing routing infrastructures. NetFlow switching is described in Chapter 8, "Configuring NetFlow Switching."
Platform and Switching Path Correlation
Depending on the routing platform you are using, availability and default implementations of switching paths varies. Table 1-1 shows the correlation between Cisco IOS switching paths and routing platforms.
| Switching Path | Cisco 7200 | Cisco 7500 | Comments | Configuration Command |
|---|---|---|---|---|
| Process switching | Yes | Yes | Initializes switching caches | no protocol route-cache |
| Fast switching | Yes | Yes | Default (except for IP) | protocol route-cache |
| Distributed switching | No | Yes | Using second-generation VIP line cards | protocol route-cache distributed |
| NetFlow switching | Yes | Yes | Configurable per interface | protocol route-cache flow |