Ethernet Switching & MAC Sublayer

Ethernet Switching Basics

• Explains Ethernet functionality in a switched network.
• Relevant topics: Ethernet frames, MAC addresses, address tables, switch speeds, and forwarding.

Ethernet Frame

• Operates in the data link and physical layers.
• Aligns with IEEE 802.2 and 802.3 standards.

Data Link Sublayers

• The 802 LAN/MAN standards, like Ethernet, contain two sublayers.
• LLC Sublayer: Uses IEEE 802.2 to specify which network layer protocol a frame uses.
• MAC Sublayer: Adheres to IEEE 802.3, 802.11, or 802.15, handling data encapsulation, media access control, and data link layer addressing.

MAC Sublayer Essentials

• Data encapsulation and media access are its responsibilities.
• IEEE 802.3 data encapsulation includes the frame structure, addressing, and error detection.
• Addressing involves source and destination MAC addresses for LAN communication.
• Error detection uses a frame check sequence (FCS) trailer.
• The MAC sublayer follows IEEE 802.3 specifications.
• Specifications vary based on media like copper or fiber.

Media Access & Legacy Ethernet

• In legacy setups with bus topology or hubs, Ethernet used a shared, half-duplex medium, which manages access through CSMA/CD.
• Modern Ethernet LANs use full-duplex switches, removing the need for CSMA/CD.

Ethernet Frame Size

• Frame size has a range: 64 bytes minimum, 1518 bytes maximum.
• The preamble field is not considered when determining frame size.
• Frames shorter than 64 bytes are considered "collision fragments" or "runt frames" and are discarded automatically.
• Frames exceeding 1500 bytes are "jumbo" or "baby giant frames".
• The frame is dropped if the transmitted frame size goes below the minimum or exceeds the maximum.
• Dropped frames often result from unwanted signals or collisions, marking them as invalid.

MAC Address Structure

• A 48-bit binary value is displayed as 12 hexadecimal values.
• An 8-bit byte translates to a hexadecimal range of 00 to FF.
• Leading zeroes are always shown.
• 0x precedes the hexadecimal value in documentation.
• Hexadecimal can be shown as a subscript 16 or with "H" (e.g., 73H).
• In an Ethernet LAN, MAC addressing identifies devices at OSI data link layer
• A MAC address uses 48 bits, shown in 12 hexadecimal digits, which equals 6 bytes.
• MAC addresses must be unique, requiring vendors to register with IEEE for a unique 6-hexadecimal OUI code.
• A MAC address has a 6-hexadecimal vendor OUI code, plus a 6-hexadecimal vendor-assigned value.

Frame Processing

• Ethernet headers include both Source and Destination MAC Addresses.
• NICs check destination MAC addresses against stored addresses in RAM.
• Frames are discarded if there's no match.
• Matched frames proceed to OSI layers for de-encapsulation.
• NICs accept broadcast or multicast frames from subscribed groups.
• All devices that source or receive Ethernet frames, like workstations or routers, have a MAC address and an Ethernet NIC.

MAC Addresses

• Unicast, broadcast, and multicast addresses are used for layer 2 communication.
• A unicast MAC address is exclusive to single devices.
• Address Resolution Protocol (ARP) locates IPv4 destination MAC addresses.
• Neighbor Discovery (ND) locates IPv6 destination MAC addresses.
• The source MAC address needs to be unicast only.

Broadcast MAC Addresses

• Ethernet broadcast frames reach every device on the LAN.
• Its destination MAC address is FF-FF-FF-FF-FF-FF (48 ones in binary).
• Except for the incoming port, all Ethernet switch ports get flooded. Routers don't forward it.
• For IPv4 broadcasts, the packet has an address using all ones (1s) within the host portion. All local network hosts receive and handle the said packet.

Multicast MAC Addresses

• Multicast frames are received and processed by devices in the defined multicast group.
• The encapsulated IPv4 multicast packet's destination MAC address: 01-00-5E.
• If IPv6 is encapsulated, the destination MAC address: 33-33.
• Other reserved multicast MAC addresses are used when data is not IP, such as Spanning Tree Protocol (STP).
• Flooding occurs on all Ethernet switch ports, excluding the incoming one, unless multicast snooping is active. Routers do not forward unless configured.
• Multicast addresses act as host groups and are exclusive to the destination of a packet, which requires a unicast source address.
• Multicast IP addresses need a matching multicast MAC address.

Switch Fundamentals

• Layer 2 switch relies solely on MAC addresses to make forwarding decisions.
• The switch is unaware of any protocol in the data portion.
• To make forwarding decisions, a switch consults its MAC address table.
• When a switch is powered on, the MAC address table is empty.
• A MAC address table is a content addressable memory (CAM) table.

Learning and Forwarding

• Switches assess each frame for new info.
• The switch learns by checking out: The source MAC address & the receiving port number .
• New MAC addresses are added to the MAC address table.
• Upon recognition of a prior source address, only the refresh timer is updated.
• Entries are kept for 5 minutes.
• A source MAC address on a different port is considered a new entry.

Destination MAC Addresses

• A unicast will cause the switch to seek match between frame and MAC address table.
• If it exists, the frame goes to that specified port.
• Otherwise, an unknown unicast floods all ports except for the incoming port.
• Broadcast and multicast frames also flood all ports but the incoming port.

Filtering frames

• Switches build out their MAC address table using the source MAC address of every frame.
• Frames are filtered and forwarded out a single port when the destination MAC address is known.

Frame Forwarding

• Store-and-forward switching receives the entire frame, computes the CRC, determines the outgoing interface, then forwards the frame.
• Cut-through switching forwards the frame before receiving all of it, reading only the destination address.
• A store-and-forward benefit includes checking for errors before propagation.
• Store-and-forward is used for quality of service (QoS).

Cut-Through Switching Particulars

• Under cut-through switching, switches act fastest due to not waiting for total receipt.
• Only enough frame data is buffered to read the destination's MAC address.
• All cut-through switches avoid error checking.
• Fast-forward switching has lowest latency as it forwards post address read.
• Because of fast forwarding, packets might relay with errors, with error correction being up to the destination NIC.
• Fragment-free switching balances the integrity of store-and-forward, but minimizes latency by storing, then checking the first 64 bytes for errors.

Memory Buffering

• Buffering is used to store frames before forwarding during congestion.
• Port-based memory contains frame queues that go to specific ports; frames wait transmission until all prior frames send.
• Shared memory involves switch ports sharing a buffer and allocating memory dynamically.
• Frames are dynamically linked to destination ports and transmitted.
• Shared memory buffering transmits bigger frames with less drops under asymmetric switching.

Duplex and Speed Settings

• Bandwidth and duplex settings are key to switch ports.
• They must match on both ends.
• Full-duplex allows simultaneous sending/receiving.
• Half-duplex allows one-way communication.
• Auto-negotiation finds the best settings.
• Gigabit Ethernet ports only use full-duplex.

Common Issues

• Having a duplex mismatch can lead to performance complications.
• These mismatches emerge after resetting one or both ports or forgetting to reconfigure.
• It is best to configure both Ethernet switch ports for full-duplex or enable autonegotiation on both sides.

Auto-MDIX Functionality

• Crossover or straight-through cables were once necessary based on the devices.
• A direct router-to-host now needs a crossover cable.
• Auto-MDIX auto-detects cables and sets interfaces.
• It has been automatically in use with Cisco IOS Release 12.2(18)SE or later and can be re-enabled via mdix auto.

Key Aspects from the Module

• Ethernet standards define both the Layer 2 protocols and the Layer 1 technologies, and operates via the LLC and MAC sublayers.
• Frames have a preamble, destination/source MAC addresses, and an FCS.
• MAC addresses identify devices in the OSI model.
• Addresses are 48 bits using 12 hex digits.
• The Ethernet header contains source/destination MAC addresses. They are used for unicast, broadcast, and multicast.
• Layer 2 Ethernet switch forwarding decisions use Layer 2 Ethernet MAC addresses dynamically built via source MAC address analysis.

Core Concepts

• Switch forwarding uses the destination MAC address through a MAC address table.
• Switches use store-and-forward or cut-through methods, the latter also includes fast-forward and fragment-free.
• Memory buffering includes port-based and shared memory.
• Duplex settings include full-duplex and half-duplex.

Labs

• Lab 7.1.6 explores Ethernet topology using Wireshark.
• Lab 7.2.7 highlights MAC addresses network devices.
• Lab 7.3.7 focuses on switch MAC address tables.