ethernet switching

• Ethernet operates in the data link layer and the physical layer
• Ethernet is a family of networking technologies defined in the IEEE 802.2 and 802.3 standards
• The 802 LAN/MAN standards, including Ethernet, use two separate sublayers of the data link layer to operate:

  • LLC Sublayer: Places information in the frame to identify which network layer protocol is used for the frame
  • MAC Sublayer: Responsible for data encapsulation and media access control, and provides data link layer addressing
• IEEE 802.3 data encapsulation includes:

  • Ethernet frame: Internal structure of the Ethernet frame
  • Ethernet Addressing: Includes both a source and destination MAC address
  • Ethernet Error detection: Includes a frame check sequence (FCS) trailer for error detection
• Legacy Ethernet using a bus topology or hubs is a shared, half-duplex medium

  • Ethernet LANs today use switches that operate in full-duplex, eliminating the need for access control through CSMA/CD
• The minimum Ethernet frame size is 64 bytes and the maximum is 1518 bytes

  • Frames less than 64 bytes are considered "collision fragments" or "runt frames" and are automatically discarded
  • Frames with more than 1500 bytes are considered "jumbo" or "baby giant frames"
• An Ethernet MAC address consists of a 48-bit binary value, expressed using 12 hexadecimal values

  • A MAC address is 6 bytes in length and must be unique to the Ethernet device or interface
• When a device forwards a message to an Ethernet network, the Ethernet header includes a Source MAC address and a Destination MAC address

  • NICs will accept frames if the destination MAC address is a broadcast or multicast group
• A unicast MAC address is used for single transmitting device to single destination device communication

  • Address Resolution Protocol (ARP) is used to determine the destination MAC address associated with an IPv4 address
• An Ethernet broadcast frame is received and processed by every device on the Ethernet LAN

  • Broadcast frames have a destination MAC address of FF-FF-FF-FF-FF-FF in hexadecimal
• An Ethernet multicast frame is received and processed by a group of devices that belong to the same multicast group

  • Multicast MAC addresses represent a group of addresses and require a corresponding multicast IP address
• A Layer 2 Ethernet switch uses MAC addresses to make forwarding decisions

  • Switches learn MAC addresses by examining the source MAC address of frames
• If the destination MAC address is a unicast address, the switch will look for a match in its MAC address table:

  • If the address is found, the switch forwards the frame out the specified port.
  • If the address is not found, the switch forwards the frame out all ports except the incoming port (unknown unicast).
  • Broadcast or multicast addresses are also flooded out all ports except the incoming port
• Switches build their MAC address tables by examining the source MAC address of received frames
• Switches use store-and-forward or cut-through switching methods for data forwarding between network ports:

  • Store-and-forward: receives the entire frame, computes CRC, looks up destination address, and forwards if valid
  • Cut-through: forwards the frame before it is entirely received, reading at least the destination address before forwarding
• Store-and-forward switching:

  • Determines frame errors before propagation, discarding frames with errors to reduce bandwidth consumption
  • Required for quality of service (QoS) analysis in converged networks for traffic prioritization, e.g., VoIP
• Cut-through switching:

  • Forwards data upon reception, buffering enough to read the destination MAC address
  • Two variants: Fast-forward (low latency, may forward faulty packets) and Fragment-free (error check on first 64 bytes before forwarding)
• Memory buffering on switches:

  • Used to store frames before forwarding or when destination port is congested
  • Shared memory buffering results in larger frames with fewer dropped frames, important for asymmetric switching
• Duplex and speed settings on switches:

  • Basic settings include bandwidth ("speed") and duplex settings for each port
  • Duplex settings: Full-duplex (send and receive simultaneously) and Half-duplex (one end sends at a time)
  • Autonegotiation enables devices to automatically negotiate speed and duplex capabilities
  • Gigabit Ethernet ports operate only in full-duplex
• Duplex mismatch:

  • Common cause of performance issues on 10/100 Mbps Ethernet links
  • Occurs when one port operates at half-duplex while the other operates at full-duplex
  • Best practice is to configure both switch ports as full-duplex
• Auto-MDIX:

  • Feature on switches that automatically detects cable type and configures interfaces accordingly
  • Enabled by default on most switches, but can be disabled
  • Best practice is to use correct cable type and not rely solely on auto-MDIX