Ethernet Switching

The principles of Ethernet switches and their operation within bridged Local Area Networks (LANs) are defined in the IEEE 802.1D standard. This standard focuses on ensuring efficient and reliable data forwarding, minimizing collisions, and enabling network scalability through bridging and switching mechanisms.

An Ethernet switch operates as a multi-port bridge that forwards Ethernet frames between different segments of a network. It is designed to improve performance and segment traffic in a LAN by making intelligent forwarding decisions based on MAC addresses. Figure 1 shows a switch with three ports and an end station connected to each port. At the same time, each station sends a frame. In a shared-medium Ethernet, collisions would occur in this case and the frames would have to be retransmitted. The switch, however, has internal memory where the frames can be buffered. Thus, Frame 1 and Frame 2, which are destined to the same station, will not collide and one of them will be stored in the buffer of the switch until the output port of the switch, where the destination station is connected, is free. Moreover, due to full duplex communication, Frame 2 and Frame 3 can be transmitted at the same time.

Using switches, the Ethernet subnet is splitted into several point-to-point connections, whereas each connection represents two interconnected ports – either between an end station and a switch, or between two switches. This allows for scaling the network by connecting additional switches.

Key principles of switch operation

1. Learning MAC Addresses:
   • Ethernet switches maintain a MAC address table (Forwarding Database), which maps each connected device’s MAC address to a specific switch port.
   • When a frame arrives at a port, the switch inspects the source MAC address and associates it with that port in its table.
   • This learning process enables the switch to efficiently forward frames to the correct destination port.
2. Forwarding and Filtering:
   • When a frame arrives, the switch checks the destination MAC address:
     • If the destination address is in the table, the frame is forwarded only to the corresponding port (unicast forwarding).
     • If the address is unknown or the frame is a broadcast/multicast, it is sent to all ports except the one it originated from (flooding).
3. Loop Prevention with Spanning Tree Protocol (STP):
   • Ethernet switches use Spanning Tree Protocol (STP), defined in IEEE 802.1D, to prevent loops in the network.
   • STP ensures that there is only one active path between any two devices by blocking redundant paths and creating a loop-free topology.
   • If a primary path fails, STP dynamically reconfigures to activate a backup path.
4. Segmentation and Collision Domain Isolation:
   • Each port on a switch represents a separate collision domain, which eliminates collisions typical in hub-based networks.
   • This isolation improves network performance by allowing multiple simultaneous transmissions.
5. Store-and-Forward Switching:
   • Switches store the entire frame before forwarding it. This allows the device to verify the integrity of the frame using the Cyclic Redundancy Check (CRC) and discard corrupted frames.

Switches are transparent, i.e. the end stations are unaware of the presence of the bridge.

The switch transparency is elucidated in Figure 2. The forwarding table of the central switch contains destination ports of the stations in the network regardless the station is connected directly to a port of the switch, or is accessible through an additional subnet. Such a subnet may be composed from a single switch or a series of switches.

Cut-through vs. Store-and-forward

Store-and-forward

• The entire frame is stored, its CRC is checked and the frame is forwarded to the destination port
• Corrupted frames are not forwarded

Cut-through

• Only a small fraction of the frame, which is needed to determine the destination port, is stored in the switch before the frame is forwarded
• Small forwarding delay of the switch
• Corrupted frames are forwarded, because the corruption cannot be detected, when the switch starts forwarding the frame
• Same speed at both connected ports (node – switch, switch – switch)