Types of Cisco Switches
Switches are the fundamental interconnect component of the campus network. Cisco offers a variety of switches specifically designed for different functions. At the time of this writing, Cisco designs the Catalyst switches for campus networks and Nexus switches for data centers. In the context of CCNP, this book focuses mostly on Catalyst switches.
Figure 2-12 illustrates the current recommended Catalyst switches. However, in the competitive campus switch marketplace, Cisco continuously updates the Catalyst switches with new capabilities, higher performance, higher density, and lower cost.
)
Figure 2-12 Cisco Catalyst Switches
Interesting enough, the Catalyst 6500 was not detailed in Figure 2-12. Despite its extremely long life cycle, Cisco marketing has finally shifted focus to the Catalyst 6800. For a large number of you reading this book, you have likely come across the Catalyst 6500 at some point in your career.
Cisco offers two types of network switches: fixed configuration and modular switches. With fixed configuration switches, you cannot swap or add another module, like you can with a modular switch. In enterprise access layers, you will find fixed configuration switches, like the Cisco Catalyst, 2960-X series. It offers a wide range of deployments.
In the enterprise distribution layer, you will find either fixed or modular switches depending on campus network requirements. An example of a modular switch that can be found in the distribution layer is the Cisco Catalyst 3850-X series. This series of switches allows you to select different network modules (Ethernet or fiber optic) and redundant power supply modules. In small businesses without a distribution layer, the 3850-X can be found in the core layer. In large enterprise networks, you might find 3850-X in the access layer in cases where high redundancy and full Layer 3 functionality at the access layer are requirements.
In the enterprise core layer, you will often find modular switches such as the Cisco Catalyst 6500 or the Catalyst 6800 series. With the 6800 switch, nearly every component, including the route processing/supervisor module and Ethernet models to power supplies) is individually installed in a chassis. This individualization allows for customization and high-availability options when necessary.
If you have a network where there is a lot of traffic, you have the option to leverage the Cisco Catalyst 4500-X series switches into the distribution layer. The Catalyst 4500-X supports supervisor/route process redundancy and supports 10 Gigabit Ethernet.
All switches within the 2960-X, 3850-X, 4500-X, and 6800-X series are managed. This means that you can configure an IP address on the device. By having a management IP address, you can connect to the device using Secure Shell (SSH) or Telnet and change device settings. An unmanaged switch is only appropriate for a home or very small business environment. It is highly recommended not to use an unmanaged switch in any campus network.
This section just described a few examples of Cisco switches and their placement in the network. For more information, go to http://www.cisco.com/c/en/us/products/switches/index.html.
The next section compares Layer 2 and Layer 3 (multilayer switches).
Comparing Layer 2 and Multilayer Switches
A Layer 2 Ethernet switch operates at the Data Link Layer of the OSI model. These types of switches make decisions about forwarding frames based on the destination MAC addresses found within the frame.
Recalling basic networking: A switch collision domain is only port to port because each switch port and its associated end device is its own collision domain. Because there is no contention on the media, all hosts can operate in full-duplex mode, which means that they can receive and transmit data at the same time. The concept of half duplex is legacy and applies only to hubs and older 10/100-Mbps switches, because 1 Gbps operates by default at full duplex.
When a switch receives in store-n-forward mode, the frame is checked for errors, and frames with a valid cyclic redundancy check (CRC) are regenerated and transmitted. Some models of switches, mostly Nexus switches, opt to switch frames based only on reading the Layer 2 information and bypassing the CRC check. This bypass, referred to as cut-through switching, lowers the latency of the frame transmission as the entire frame is not stored before transmission to another port. Lower switching latency is beneficial for low-latency applications such as algorithm trading programs found in the data center. The assumption is that the end device network interface card (NIC) or an upper-level protocol will eventually discard the bad frame. Most Catalyst switches are store-n-forward.
MAC Address Forwarding
To figure out where a frame must be sent, the switch will look up its MAC address table. This information can be told to the switch or it can learn it automatically. The switch listens to incoming frames and checks the source MAC addresses. If the address is not in the table already, the MAC address, switch port, and VLAN will then get recorded in the forwarding table. The forwarding table is also called the CAM table.
What happens if the destination MAC address of the frame is unknown to the switch? The switch then forwards the frame through all ports within a VLAN except the port the frame was received on. This is known as unknown unicast flooding. Broadcast and multicast traffic is destined for multiple destinations, so it will get flooded by default.
Referring to Figure 2-13, in the first example, the switch receives a frame on port 1. The destination MAC address for the frame is 0000.0000.5555. The switch will look up its forwarding table and figure out that MAC address 0000.0000.5555 is recorded on port 5. The switch will then forward the frame through port 5.
)
Figure 2-13 Layer 2 Switching Operation: MAC Address Forwarding
In the second example, the switch receives a broadcast frame on port 1. The switch will forward the frame through all ports that are within the same VLAN except port 1. The frame was received on port 1, which is in VLAN 1; therefore, the frame is forwarded through all ports on the switch that belong to VLAN 1 (all ports except port 3).
The next subsection discusses Layer 2 switch operation from a mechanics point of view.
Layer 2 Switch Operation
When a switch receives a frame, it places the frame into an ingress queue. A port can have multiple ingress queues, and typically these queues are used to service frames differently (for example, apply quality of service [QoS]). From a simplified viewpoint, when the switch selects a frame from a queue to transmit, the switches need to answer a few questions:
- Where should the frame be forwarded?
- Are there restrictions preventing the forwarding of the frame?
- Is there any prioritization or marking that needs to be applied to the frame?
Decisions about these three questions are answered, respectively, as illustrated in Figure 2-14 and described in the list that follows.
)
Figure 2-14 Layer 2 Switch Operation: Mechanics
- Layer 2 forwarding table: The Layer 2 forwarding table, also called the MAC table, contains information about where to forward the frame. Specifically, it contains MAC addresses and destination ports. The switches reference the destination MAC address of the incoming frame in the MAC table and forward the frames to the destination ports specified in the table. If the MAC address is not found, the frame is flooded through all ports in the same VLAN.
- ACLs: Access control lists (ACLs) do not only apply to routers. Switches can also apply ACLs based on MAC and IP addresses. Generally only higher-end switches support ACLs based on both MAC and IP addresses, whereas Layer 2 switches support ACLs only with MAC addresses.
- QoS: Incoming frames can be classified according to QoS parameters. Traffic can then be marked, prioritized, or rate-limited.
Switches use specialized hardware to house the MAC table, ACL lookup data, and QoS lookup data. For the MAC table, switches use content-addressable memory (CAM), whereas the ACL and QoS tables are housed in ternary content-addressable memory (TCAM). Both CAM and TCAM are extremely fast access and allow for line-rate switching performance. CAM supports only two results: 0 or 1. Therefore, CAM is useful for Layer 2 forwarding tables.
TCAM provides three results: 0, 1, and don’t care. TCAM is most useful for building tables for searching on longest matches, such as IP routing tables organized by IP prefixes. The TCAM table stores ACL, QoS, and other information generally associated with upper-layer processing. As a result of using TCAM, applying ACLs does not affect the performance of the switch.
This section only touches on the details and implementation of CAM and TCAM needed for the CCNP certification. For a more detailed description, review the following support document at Cisco.com:
The next subsection discusses Layer 3 (multilayer) switch operation in more detail.
Layer 3 (Multilayer) Switch Operation
Multilayer switches not only perform Layer 2 switching but also forward frames based on Layer 3 and 4 information. Multilayer switches not only combine the functions of a switch and a router but also add a flow cache component.
Multilayer switches apply the same behavior as Layer 2 switches but add an additional parallel lookup for how to route a packet, as illustrated in Figure 2-15.
)
Figure 2-15 Multilayer Switch Operation
The associated table for Layer 3 lookups is called a FIB table. The FIB table contains not only egress ports and VLAN information but also MAC rewrite information. The ACL and QoS parallel lookups happen the same as Layer 2 switches, except there may be additional support for Layer 3 ACLs and QoS prioritization.
For example, a Layer 2 switch may only be able to apply to rate-limiting frames based on source or destination MAC addresses, whereas a multilayer switch generally supports rate-limiting frames on IP/MAC addresses.
Unfortunately, different models of Cisco switches support different capabilities, and some Layer 2-only switches actually support Layer 3 ACLs and QoS lookups. It is best to consult the product documentation at Cisco.com for clear information about what your switch supports. For the purpose of CCNP Switch and the context of this book, Layer 2 switches support ACLs and QoS based on MAC addresses, whereas Layer 3 switches support ACLs and QoS based on IP or MAC addresses.
Useful Commands for Viewing and Editing Catalyst Switch MAC Address Tables
There is one command for viewing the Layer 2 forwarding table on Catalyst and Nexus switches: show mac address-table. The table has many optional parameters to narrow the output to a more manageable result in large networks. The full command options are as follows: show mac-address-table [aging-time | count | dynamic | static] [address hw-addr] [interface interface-id] [vlan vlan-id] [ | {begin | exclude | include} expression].
Example 2-1 illustrates sample uses of the command and several useful optional uses.
Example 2-1 Layer 2 Forwarding Table
Switch1# show mac address-table
Mac Address Table
-------------------------------------------
Vlan Mac Address Type Ports
---- ----------- -------- -----
1 0000:0c00.9001 DYNAMIC Et0/1
1 0000.0c00.9002 DYNAMIC Et0/2
1 0000.0c00.9002 DYNAMIC Et0/3
Total Mac Addresses for this criterion: 3
Switch1# show mac address-table interface ethernet 0/1
Mac Address Table
-------------------------------------------
Vlan Mac Address Type Ports
---- ----------- -------- -----
1 0000:0c00.9001 DYNAMIC Et0/1
Total Mac Addresses for this criterion: 1
Switch1# show mac address-table | include 9001
1 0000:0c00.9001 DYNAMIC Et0/1
Frame Rewrite
From your CCNA studies, you know that many fields of a packet must be rewritten when the packets are routed between subnets. These fields include both source and destination MAC addresses, the IP header checksum, the TTL (Time-to-Live), and the trailer checksum (Ethernet CRC). See Chapter 1, “Fundamentals Review,” for an example.
Distributed Hardware Forwarding
Network devices contain at least three planes of operation:
- Management plane
- Control plane
- Forwarding plane
The management plane is responsible for the network management, such as SSH access and SNMP, and may operate over an out-of-band (OOB) port. The control plane is responsible for protocols and routing decisions, and the forwarding plane is responsible for the actual routing (or switching) of most packets.
Multilayer switches must achieve high performance at line rate across a large number of ports. To do so, multilayer switches deploy independent control and forwarding planes. In this manner, the control plane will program the forwarding plane on how to route packets. Multilayer switches may also employ multiple forwarding planes. For example, a Catalyst 6800 uses forwarding planes on each line module, with a central control plane on the supervisor module.
To continue the example of the Catalyst 6800, each line module includes a microcoded processor that handles all packet forwarding. For the control plane on the supervisor to communicate with the line module, a control layer communication protocol exists, as shown in Figure 2-16.
)
Figure 2-16 Distributed Hardware Forwarding
The main functions of this control layer protocol between the control plane and the forwarding plane are as follows:
- Managing the internal data and control circuits for the packet-forwarding and control functions
- Extracting the other routing and packet-forwarding-related control information from the Layer 2 and Layer 3 bridging and routing protocols and the configuration data, and then conveying the information to the interface module for control of the data path
- Collecting the data path information, such as traffic statistics, from the interface module to the route processor
- Handling certain data packets that are sent from the Ethernet interface modules to the route processor (for example, DCHP requests, broadcast packets, routing protocol packets)
Cisco Switching Methods
The term Cisco switching methods describes the route processor behavior found on Cisco IOS routers. Because multilayer switches are capable of routing and, in fact, contain a routing process, a review of these concepts is necessary.
A Cisco IOS-based router uses one of three methods to forward packets: process switching, fast switching, and Cisco Express Forwarding (CEF). Recall from your study of routers that process switching is the slowest form of routing because the router processor must route and rewrite using software. Because speed and the number of cores limit the route processor, this method does not scale. The second method, fast switching, is a faster method by which the first packet in a flow is routed and rewritten by a route processor using software, and each subsequent packet is then handled by hardware. The CEF method uses hardware forwarding tables for most common traffic flows, with only a few exceptions. If you use CEF, the route processor spends its cycles mostly on other tasks.
The architecture of the Cisco Catalyst and Nexus switches both focus primarily on the Cisco router equivalents of CEF. The absolute last-resort switching method for Cisco Catalyst or Nexus switches is process switching. The route processors of these switches were never designed to switch or route packets, and by doing so, this will have an adverse effect on performance. Fortunately, the default behavior of these switches is to use fast switching or CEF, and process switching occurs only when necessary.
With Cisco Catalyst switching terminology, fast switching is referred to as route caching, and the application of CEF with distributed hardware forwarding is referred to as topology-based switching.
As a review, the following list summarizes route caching and topology-based forwarding on Cisco Catalyst switches:
- Route caching: Also known as flow-based or demand-based switching, route caching describes a Layer 3 route cache that is built within the hardware functions as the switch detects traffic flow into the switch. This method is functionally equivalent to fast switching in Cisco IOS Software.
- Topology-based switching: Information from the routing table is used to populate the route cache, regardless of traffic flow. The populated route cache is the FIB, and CEF is the facility that builds the FIB. This method is functionally equivalent to CEF in Cisco IOS Software.
The next subsections describe route caching and topology-based switching in more detail.
Route Caching
Route caching is the fast switching equivalent in Cisco Catalyst switches. For route caching to operate, the destination MAC address of an incoming frame must be that of a switch interface with Layer 3 capabilities. The first packet in a stream is switched in software by the route processor, because no cache entry exists yet for the new flow. The forwarding decision that is made by the route processor is then programmed into a cache table (the hardware forwarding table), and all subsequent packets in the flow are switched in the hardware, commonly referred to as using application-specific interface circuits (ASICs). Entries are created only in the hardware forwarding table as the switch detects new traffic flows, and entries will time out after they have been unused for a period of time.
Because entries are created only in the hardware cache as flows are detected by the switch, route caching will always forward at least one packet in a flow using software.
Route caching carries many other names, such as NetfFow LAN switching, flow-based or demand-based switching, and route once, switch many.
Figure 2-17 briefly highlights this concept from a hardware perspective.
)
Figure 2-17 Route Caching
Topology-Based Switching
Topology-based switching is the CEF equivalent feature of Cisco Catalyst switches. Topology-based switching is ideal for Layer 3 switching over route caching because it offers the best performance and scalability. Fortunately, all Cisco Catalyst switches capable of Layer 3 routing leverage topology-based switching / CEF. For the purpose of CCNP Switch, focus primarily on the benefits and operation of topology-based switching.
CEF uses information in the routing table to populate a route cache (known as an FIB), without traffic flows being necessary to initiate the caching process. Because this hardware FIB exists regardless of traffic flow, assuming that a destination address has a route in the routing table, all packets that are part of a flow will be forwarded by the hardware. The FIB even handles the first packet of a flow. Figure 2-18 illustrates this behavior.
)
Figure 2-18 Topology-Based Switching
In addition, CEF adds enhanced support for parallel paths and thus optimizes load balancing at the IP layer. In most current-generation Catalyst switches, such as the Catalyst 4500 and 6800, CEF supports both load balancing based on source IP address and destination IP address combination and source and destination IP plus TCP/UDP port number.
CEF load-balancing schemes allow for Layer 3 switches to use multiple paths to achieve load sharing. Packets for a given source-destination host pair are guaranteed to take the same path, even if multiple paths are available. This ensures that packets for a given host pair arrive in order, which in some cases may be the desired behavior with legacy applications.
Moreover, load balancing based only on source and destination IP address has a few shortcomings. Because this load-balancing method always selects the same path for a given host pair, a heavily used source-destination pair, such as a firewall to web server, might not leverage all available links. In other words, the behavior of this load-balancing scheme may “polarize” the traffic by using only one path for a given host pair, thus effectively negating the load-balancing benefit of the multiple paths for that particular host pair.
So, optimal use of any load-balancing scheme depends on the statistical distribution of traffic because source and destination IP load sharing becomes more effective as the number of source-destination IP pairs increases. In an environment where there is a broad distribution of traffic among host pairs, polarization is of minimal concern. However, in an environment where the data flow between a small number of host pairs creates a disproportionate percentage of the packets traversing the network, polarization can become a serious problem.
A popular alternative that is now the default behavior in new Catalyst switches is load balancing based on source and destination IP to include TCP/UDP port numbers. The more additional factors added to the load-balancing scheme, the less likely polarization will exist.
Cisco Catalyst supports additional load-balancing methods and features by which to tune load balancing based on hardware model and software version. Consult Cisco.com for such configuration optimizations if necessary.
Hardware Forward Details
The actual Layer 3 switching of packets occurs at two possible different locations on Catalyst switches. These possible locations are in a centralized manner, such as on a supervisor module, or in distributed fashion, where switching occurs on individual line modules. These methods are referred to as centralized switching and distributed switching, respectively.
The Catalyst 6500 was a perfect example where there was an option to centralize switch everything on the supervisor or place specific hardware versions of line modules in the chassis to gain distributed switching capability.
The benefits of centralized switching include lower hardware cost and lower complexity. For scaling and large enterprise core networks, distributed switching is optimal. Most small form-factor switches leverage centralized switching.
In conclusion, the subsections of this chapter pertaining to switching methods and hardware forwarding included many specific details about routing and switching operations on Cisco switches. Among all the lengthy explanations and details, conclude this section with the following concepts:
- The control plane (CPU/route processor) of a Cisco Catalyst was never designed to route or switch frames. The control plane is intended only to populate hardware tables with routing information and maintain routing protocols. The control plane may route frames in a few exception conditions.
- Medium- to high-end Cisco Catalyst switches were designed based on the distributing forward model to scale to demands of campus and data center networks.
- Cisco Catalyst switches leverage CEF (topology-based switching) for routing of frames as a means to implement a distributing hardware forwarding model.
- Cisco Catalyst switches use either a centralized method or a distributed line module method of hardware forwarding, depending on specific platform model and configuration.