Ethernet Switching and IEEE 802.3 standards

Module 7: Ethernet Switching

• Explains how Ethernet works in a switched network.

Ethernet Frames

• Ethernet operates in the data link and physical layers.
• It is a family of networking technologies defined in IEEE 802.2 and 802.3 standards.
• The 802 LAN/MAN standards use two separate sublayers of the data link layer: LLC and MAC.
• LLC Sublayer (IEEE 802.2) includes information in the frame to identify the network layer protocol.
• MAC Sublayer (IEEE 802.3, 802.11, or 802.15) is responsible for data encapsulation, media access control, and data link layer addressing.
• The MAC sublayer is responsible for data encapsulation and media access.
• IEEE 802.3 data encapsulation includes Ethernet frame structure, addressing, and error detection.
• The IEEE 802.3 MAC sublayer specifies different Ethernet communication standards over copper and fiber media.
• Legacy Ethernet (bus topology/hubs) uses a shared, half-duplex medium and the CSMA/CD access method.
• Modern switched Ethernet LANs operate in full-duplex and don't require CSMA/CD.
• Ethernet frame's minimum size is 64 bytes, maximum is 1518 bytes (excluding preamble).
• Frames under 64 bytes are "collision fragments" or "runt frames" that are discarded.
• Frames over 1500 data bytes are "jumbo" or "baby giant" frames.
• Receiving device will drop frames smaller or larger than specified size, as result of collisions/unwanted signals and are considered invalid.
• Jumbo frames are supported by most Fast Ethernet and Gigabit Ethernet switches and NICs.
• Wireshark can be used to examine header fields in Ethernet II frames and to capture and analyze Ethernet frames.

Ethernet MAC Address

• An Ethernet MAC address is a 48-bit binary value, shown as 12 hexadecimal values.
• 8 bits (one byte) can be represented in hexadecimal as the range 00 to FF.
• When using hexadecimal, leading zeroes are always displayed to complete the 8-bit representation.
• Hexadecimal numbers are often preceded by "0x" (e.g., 0x73) in documentation.
• Hexadecimal numbers may be represented by a subscript 16 or followed by "H" (e.g., 73H).
• Every network device in an Ethernet LAN is connected to shared media; MAC addresses provide identification at the data link layer of the OSI model.
• A MAC address is 48 bits, or 6 bytes, expressed using 12 hexadecimal digits.
• Each Ethernet device/interface needs a unique MAC address; vendors register with the IEEE for a unique 24-bit (3-byte) code (OUI - organizationally unique identifier).
• An Ethernet MAC address is a 6 hexadecimal vendor OUI code followed by a 6 hexadecimal vendor-assigned value.
• When forwarding a message, the Ethernet header includes source and destination MAC addresses.
• When a NIC receives an Ethernet frame, it checks to see if the destination MAC address matches its physical MAC address stored in RAM.
• Frames with a mismatched MAC address are discarded while frames with matching MAC addresses are passed up the OSI layers for de-encapsulation.
• Ethernet NICs accept frames addressed to a broadcast or multicast group to which they belong.
• All devices (workstations, servers, printers, mobile devices, routers) that are source or destination of Ethernet frames have Ethernet NICs and MAC addresses.
• In Ethernet, different MAC addresses are used for Layer 2 unicast, broadcast, and multicast communications.
• A unicast MAC address is unique and used for one-to-one communication from a single transmitting device.
• Address Resolution Protocol (ARP) is the process a source host uses to determine the destination MAC address associated with an IPv4 address.
• Neighbor Discovery (ND) is the process a source host uses to determine the destination MAC address associated with an IPv6 address.
• The source MAC address must always be a unicast address.
• An Ethernet broadcast frame is received and processed by every device on the Ethernet LAN.
• A broadcast has a destination MAC address of FF-FF-FF-FF-FF-FF (48 ones in binary).
• Broadcasts are flooded out of all Ethernet switch ports except the incoming port but are not forwarded by a router.
• IPv4 packets in an Ethernet broadcast frame contain a destination IPv4 address with all ones (1s) in the host portion so all of the hosts will receive and process it.
• An Ethernet multicast frame is received and processed by devices that belong to the same multicast group.
• A destination MAC address of 01-00-5E indicates an IPv4 multicast packet, and 33-33 indicates an IPv6 multicast packet.
• Multicasts are flooded out all Ethernet switch ports (except the incoming port, unless multicast snooping is configured).
• It is not forwarded by a router, unless the router is configured to route multicast packets.
• Multicast addresses represent a group of addresses and are used as the packet destination while the source is always a unicast address.
• Like unicast and broadcast, multicast IP addresses need a corresponding multicast MAC address.
• You can set up the topology, configure devices, verify connectivity, display, describe and analyze Ethernet MAC addresses.

The MAC Address Table

• Layer 2 Ethernet switches use Layer 2 MAC addresses to make forwarding decisions.
• Ethernet switches are not aware of higher-layer protocol data (IPv4 packets, ARP messages, IPv6 ND packets).
• An Ethernet switch examines its MAC address table to make a forwarding decision for each frame, unlike legacy Ethernet hubs.
• The MAC address table is empty when a switch is turned on.
• MAC address table is sometimes referred to as a content addressable memory (CAM) table.
• Switches check every incoming frame for new information by examining the source MAC address and the port the frame entered on.
• If the source MAC address doesn't exist, it gets added to the table along with the incoming port number, otherwise, if that source MAC address does exist, the refresh timer for the entry gets updated.
• Ethernet switches keep an entry in the table for 5 minutes by default unless it is reset.
• When the source MAC address exists but is on a different port, it is treated as a new entry and replaces the old one.
• For destination MAC addresses in a unicast frame, there is a look-up for a match between the destination MAC address of the frame and an entry in the MAC address table.
• If the destination MAC address is in table, the frame gets forwarded out that specified port.
• If the destination MAC address is not in the table, the switch will forward (flood)the frame out all ports except that incoming port.
• This is called an unknown unicast.
• If the destination MAC is a broadcast or multicast, the frame floods to all ports, except the incoming.
• Switches populate the MAC address table by examining the source MAC address of every frame.
• They are able to filter the frame and forward out a single port when the destination MAC address is contained in the MAC address table.

Switch Speeds and Forwarding Methods

• Switches use store-and-forward or cut-through forwarding methods to switch data between network ports.
• Store-and-forward switching - receive the entire frame, compute the CRC, and if valid, look up the destination address, which determines the outgoing interface.
• Cut-through switching - forward the frame before fully receiving it. At a minimum, read the destination address before forwarding.
• The advantage of store-and-forward switching is that it checks for errors before propagating a frame; when errors are detected, the frame is discarded.
• Discarding frames with errors reduces consumed bandwidth.
• Store-and-forward switching is required for quality of service (QoS) analysis on converged networks for traffic prioritization (e.g., VoIP).
• In cut-through switching the switch acts upon the data as soon as it is received, even if the transmission is not complete.
• Switch buffers just enough of the frame to read the destination MAC address to determine which port to forward the data to.
• The switch does not perform any error checking on a cut-through frame.
• Fast-forward switching has the lowest latency, immediately forwarding a packet after reading the destination address.
• Times exist when packets are relayed with errors, because fast-forward switching starts forwarding before the entire packet has been received.
• Fragment-free switching is a compromise between store-and-forward (high latency and integrity) and fast-forward (low latency and reduced integrity).
• The switch stores and error-checks the first 64 bytes before forwarding (most network errors and collisions occur during the first 64 bytes).
• Ethernet switches may use buffering to store frames when forwarding or when the destination is busy.
• Port-based buffering: Frames are stored in queues linked to specific incoming and outgoing ports. A frame transmits to the outgoing port only upon successful transmission of all frames ahead in queue.
• Frames being delayed in a single frame causes a delay the transmission of all of the frames in the memory because of a busy destination port.
• Shared memory buffer: common memory buffer for all switch ports that allocates memory dynamically as required by packets.
• Allows receiving a packet on one port and transmitting on another without moving it to a different queue.
• Results in larger frames being transmitted with fewer dropped frames, and also provides asymmetric switching, which allows for more bandwidth dedicated to certain ports for data rates
• Two basic switch settings are bandwidth ("speed") and duplex; they should match the settings of connected devices.
• There are two duplex types for Ethernet networks: full-duplex and half-duplex.
• Full-duplex is when both ends can send and receive simultaneously.
• Half-duplex - only one can send at a time.
• The autonegotiation enables two devices to automatically negotiate the best speed and duplex capabilities.
• Gigabit Ethernet ports only operate in full-duplex.
• Duplex mismatch can cause performance issues on 10/100 Mbps Ethernet links and occurs when one port operates at half duplex while the other port operates as full-duplex.
• Resetting ports and auto-negotiation process may cause for link partners to not have the same configuration.
• Both sides of a link can be configured with autonegotiation on, or both sides can have autonegotiation off.
• Configuring both Ethernet switch ports as full-duplex is common practice.
• Auto-MDIX enabled devices to connect to other devices with either a crossover or straight-through cable.
• Required cable depends on the type of interconnecting devices.
• Most switch devices support the automatic medium-dependent interface crossover (auto-MDIX) feature to automatically detect the type of cable attached to a port and configure interfaces accordingly.
• Enabled the switch automatically detects the type of cable attached to the port and configures the interfaces accordingly.
• Runs on Cisco IOS Release 12.2(18)SE or later by default, you should always use the correct cable type and not rely on the auto-MDIX feature.
• Auto-MDIX can be re-enabled using the mdix auto interface configuration command.