Ethernet Switching Fundamentals

Ethernet Switching Fundamentals : After the Ethernet network structure, Technopediasite.com is going to provide complete information about Ethernet switching fundamentals. To get the correct and complete information it is necessary to read any article carefully till the end. Thanks to all of you who have liked the Ethernet network structure article. Now I am going to talk about Ethernet switching.

Today’s Ethernet networks are implemented through dedicated point-to-point connections. The transmission of Ethernet packets is possible in half and full duplex whereby full duplex transmission prevails the most. The switching of the Ethernet packets is based on the Store & Forward Algorithm.

Switching - Dedicated Links for end users

I have already mentioned above that Ethernet Networks  dedicated point-to-point connections and transmission of Ethernet packet in half and full duplex. Now we are also familiar that switching of the Ethernet packets is based on the Store & Forward Algorithm.

At the input of the switch, the packet is buffered. Depending on its identification (destination address), the packet is switched to the output buffer via which the B user can be reached. For the user, the connection always appears to exist since the packets are transmitted to the desired objective if any are available.

In reality, however, the connection is not through-connected; it only exists only during the delay interval of the packets. Here, we speak of a ”virtual connection”. Here, the packet stands for a data packet or an ATM cell.

Packet-switched entails:

Flexible transmission bandwidths. The bandwidth is composed of a number of packets per time unit.

Many packets/s = Wide bandwidth. 

Few packets/s = Small bandwidth.

Traffic management is necessary in order to be able to guarantee users a defined

bandwidth.

Delay fluctuations: The buffering of packets results in delay fluctuations.

Repeaters/Hubs - Work at Layer 1 of OSI Model

Repeaters or Hubs are used to extend the physical cable length of the LAN.  As its name implies, a repeater amplifies (repeats) the signal.

Layers 2 and above are passed through a hub transparently, therefore there is no link layer or network layer filtering capability.  Hubs are protocol-independent devices.

The network architecture (i.e. Ethernet, Token Ring, etc.) cannot be modified by a hub.  The Physical layer remains intact through the hub.

10Base-T and Hubs

The image shows that Layer 1 device and its Operates in Half Duplex. Here occur Collisions, retransmissions, and FCS errors are an issue. In this switching here also repeats every received frame out every other outgoing interface.

UTP made the wiring less expensive and much more flexible.  Around the same time as UTP cable for Ethernet were introduced NIC’s that combined both AUI and RJ-45 connectors. This made Ethernet explode into the enterprise as an easy and inexpensive way of setting up a network. 

 

Communication on the Layer 1 Hubs was shared and half duplex.  They were prone to collisions, retransmissions and FCS errors.  Also since the medium was shared as more devices were connected to the network each device was able to transmit less and less in the same amount of time since more stations were contending for the line but only one can use it at any given time or the results was collision of the bits. 

A major draw back to this is that all communications are seen be all devices since the Hub transmits everything entering one port to all other interfaces.  This is great for network monitoring purposes but not so much for security and efficient network communication when many devices are connected to the Hub and all trying to communicate.

Switches/Bridges - Work at Layer 2 of OSI Model

Switches or Bridges are devices that forward Data Link frames between LANs.  Layers 3 and above are passed through a switch transparently, therefore there is no network layer filtering capability.  

Switches are protocol-independent devices. Switches examine the (link layer) destination address of every frame and decide whether or not to forward the frame to the other side. This link layer filtering keeps local traffic from affecting the performance of the remote segment.

100Base-T and Switching

Switching was a big revolution in Ethernet networking. Building and expanding on the role bridges played on the network; switches were essentially multi port bridges.  Switches look like Hubs and Ethernet cables connect to them creating the Ethernet network.  

Switches no longer simply repeat all received signals out all interfaces like a Hub but instead is an intelligent device which learns where a device exists on the network in terms of what interface on the switch a device is connected to.  

It reads Ethernet frames which pass through its interfaces and creates an MAC address table which include the source MAC addresses of the Ethernet frames which have passed through it. Switches also make it possible to connect devices of different speeds to the same switch making communication between all speeds of Ethernet possible.

Switches have the capability of functioning in full duplex mode, eliminating collisions. Retransmissions because of collisions should no longer take place making communication quicker and FCS errors should be very rare on a switch.  

If FCS errors exist on switch it is probably the fault of a faulty NIC, a cable problem or some device is operating in half duplex mode, look for Hubs connected to switch ports and collisions at the same time of FCS errors.  Or one of the two directly connected devices experiencing the collisions would be in a half duplex mode and the other device in a full duplex mode. Now we can say that-

➤Don’t repeat signals out all interfaces like a hub.

➤In the switching system Intelligent device which learns where a device exists on the network.

➤Reads frames which pass through it.

➤Create a MAC address table.

➤Connects different speed devices.

➤Full duplex, and no collisions.

Switching functions, and MAC address learning

Ethernet Switches enable full duplex operation and create dynamic virtual circuits between devices connected to the network.  The way this is accomplished is by the following method.

When an Ethernet device transmits a frame it contains both the source and destination MAC address.  The switch receives this frame, verifies both the source and destination.  

It uses the source address to create a MAC address table indicating where that MAC address can be found.  Example device with MAC address 23-DC-12-56-DF-DE would be associated to the switch interface 1 in the diagram above.  The switch will read the destination MAC address, and look it up in it’s MAC address table.  

If it is located then the switch will forward the frame out the indicated switch interface.  If it is not a known address in the MAC address table then the switch will Flood (send out all interfaces) the frame.  When the end device replies the switch will then learn where the device MAC address is located because that device will include it in the source address of the reply frame.  The switch adds the MAC address into the MAC address table and then the subsequent communications will not require a flood since the MAC address will be found in the MAC address table.

Store-and-forward Technique

With the store-and-forward switching method, the LAN switch copies the entire frame to its onboard buffers and computes the cyclic redundancy check (CRC). The frame is rejected if it contains a CRC error or if it is a runt (less than 64 bytes including CRC) or a huge (1518 bytes, including CRC). 

If there is no error in the frame, the LAN switch looks at its forwarding, or switching, destination address in the table and determines the outgoing interface. It then moves the frame towards its destination.

LAN switches must use store-and-forward technologies to support multilayer switching. The switch should receive the full frame before performing any protocol-layer operation. For this reason, advanced switches that switch Layer 3 are store-and-forward devices. In short I can write below-

Phase 1: Frame detected at source port

Phase 2: The entire frame is read in 

Phase 3: Frame validity is checked (e.g. FCS)

Phase 4: Destination port is determined using destination address.

Phase 5: Frame is sent out on destination port

The latency is a function of the frame length and the internal processing time.

Cut-through Technique

Phase 1: Destination port is determined as soon as destination address is received.

Phase 2: Frame is immediately sent out on destination port without buffering and without checking.

The latency is a function only of the internal processing time.

With the cut-through switching method, the LAN switch copies only the destination address (the first 6 bytes after the preamble) to its destination buffers. It then looks at the destination address in its switching table, determines the outgoing interface, and moves the frame towards its destination. 

A cut-through switch provides low latency as it begins to forward the frame as soon as the destination address is read and determines the outgoing interface. Some switches can be configured to perform cut-through switching on a per-port basis until a user-defined error limit is reached, when they are automatically switched to store-and-forward mode. When the error rate falls below the limit, the port automatically changes back to store-and-forward mode.

Flow Control

➤In flow control used in Full Duplex  mode only. A mechanism used to control network congestion and prevent traffic from stopping.

➤If the receiver station is congested, it sends a PAUSE frame to the link partner and the link partner will stop data transmission for a short time.

➤Flow control can be enabled or disabled

- If Flow control is disabled and congestion occurs frames will be lost

➤PAUSE Frames

- Uses a defined MAC control frame format 

- Exclusive MAC address, L/T identifier

- In flow control contains a time field that defines PAUSE time

- The receipt of a PAUSE frame causes the device to stop transmission for PAUSE time.

Flow control frames cause a transmitter of data to "pause" (which prevents transmission of frame data) for a given length of time that allows the receiver to "catch up" the data it receives. And empty your received buffers.  

Switches and other flow control-aware clients use flow control.  A switch can request that the end of the link stop transmitting thereby preventing its receive buffer from overflowing and dropping frames by transmitting a PAUSE Frame and is an optional function used on full-duplex connections only.  The control frames are interspersed with normal data frames transmitted on the link.  

Let follow an example.  Station 1 is transmitting to Station 8, across the 100 Mbps Link. Then station 2 begins to transmit to station 7.  Then other stations also begin to transmit across the 100 Mbps link. When the amount of traffic begins to reach to maximum capacity of the 100 Mbps link then the switch will send PAUSE frames to the stations connected to it.  Those stations will then slow down and their total combined capacity usage will be equal or close to the 100 Mbps available. 

PAUSE frames prevent dropping of frames but the pause function does not solve the problem of steady state network congestion.  If the network is overloaded then flow control would just slow down communications so frames are not dropped but does not rectify the actual root of the problem which is probably a lack of capacity or no prioritization.  

It does not provide end to end flow control, pause frames do not cross devices, it is a frame sent interface to interface. And since it is an optional parameters some devices might have it configured on and others off, which could potentially create adverse effects.

Last Word 

Dear friends, this article Ethernet switching fundamentals is for your convenience and benefit. You should benefit from it and share it with others. Technopediasite.com tries its best to bring you new information. For any information related to Ethernet, you can contact to us. How did you like this article "Ethernet Switching Fundamentals" Please tell me by commenting.