But before we jump into the fun theories such as Spanning Tree Protocol (STP), let’s take a look at the devices we are dealing with.

Layer 2 Bridges

As you’ll recall from the OSI model, layer 2 corresponds to the Data Link layer- the layer that deals directly with MAC addresses. In this case, we are dealing with bridges that, coincidentally, handle MAC addresses.

The purpose of a bridge is simple: divide a network into two separate pieces so we can save bandwidth. That way if a computer on one segment of the bridge needs to communicate with another computer on that side of the bridge, the connection remains local. The other segment will not be bothered with the request.

This also effectively gives us two separate collision domains. This will help cut down on data collision, which is a major cause of network latency. Note that in the above example, computer A sends information to the bridge first, which makes the decision to route the request to computer B, while filtering the data out of the segment on the right.

Things to Remember About Bridges

---

• 1. Bridges provide switching via comparing destination MAC addresses found in the data being sent to MAC addresses stored in its tables.
• 2. If the source MAC address is not already known, the bridge creates a new entry in the MAC address table with the source port. This will be used for future switching operation.
• 3. If the destination MAC address is not known by the bridge, a broadcast will be sent to all segments in a process called flooding. Note that a broadcast is not sent out in the port the data was received on.
• 4. If the bridge determines the destination MAC address is not from the same network segment as the sending device, it will forward the data to the appropriate segment.

---

Layer 2 Switches

Layer 2 switches are essentially the same as bridges, only they have multiple ports and can use microsegmentation to decrease collisions and increase throughput. They also have support for full-duplex operation and spanning tree protocol (STP).

Full-duplex operation enables devices to have support for both receiving and sending information at the same time. This eliminates the problem of data collision altogether. Keep in mind that if a device such as a hub were used, full duplex operation could not be possible because hubs lack microsegmentation.

Lastly, switches use what is called spanning tree protocol. Spanning tree protocol is used to help prevent loops from forming. Imagine that switch A forwards data to switch B, since it is unsure where the data should go. Switch B isn’t sure either, and forwards the data back. This creates a never ending loop in theory, but thankfully we can make use of STP. In the example below, you can see that there are four physical links maintained, while two of the links are purely logical- they do not technically exist to the switches.

To counteract the threat of loops, switches send messages called bridge protocol data units, or BPDUs, out every port to let other switches know if its existences. Redundant paths are then shut down through port blocking, and we result with a path free of loops. (These paths can be opened again, however, in case a line goes down and the redundancy is needed.) In the example below, you can see that there are four physical links maintained, while two of the links are purely logical- they do not technically exist to the switches.

Layer 2 Switch Modes of Operation

But what sets them apart from bridges even further is that they can operate in three different modes- Store-and-Forward, Cut-Through, and Fragment-Free.

---

• 1. Store-and-Forward is the method with the slowest operation speed. This is due to the fact that it checks incoming frames of data for integrity. If the frame has errors, it is discarded. Otherwise, it is sent to its destination. This error checking can be quite costly to network performance, however.
• 2. Cut-Through switching is considered to be the bare minimum- and thus much faster. It only requires that the beginnings of the frame up to the destination MAC address be read before the frame of data can pass through the switch.
• 3. Fragment-Free switching is a modified form of the Cut-Through method. Fragment-free switching filters out collision fragments, which is where the majority of packets errors originate. To do this checking, the switch must wait for the entire packet of information to be received before the filtering takes place. Obviously, it’ll not be as quick as the cut-through method.

---

A Last Note on Broadcasts and Collisions

Keep in mind that since we have been dealing with layer 2 devices, we do not divide the broadcast domain with these devices. Only a router can divide a broadcast domain. As for collision domains, each of these devices creates more of them (this is a good thing). As for hubs and repeaters, they only extend the collision domain.

Also keep in mind that if a frame is bearing the format of FFFF.FFFF.FFFF, it will be automatically received by all NICs on the network, as this is, in fact, a broadcast address in hexadecimal.

What is a Hub?

Think of a hub as a rather naïve device- it operates on the Physical Layer of the OSI model. Switches, on the other hand, are much more intelligent- and operate on the Data Link Layer of the OSI model.

When a hub receives information on one port, the information is then broadcasted to all of its ports. This wouldn’t be such a bad plan if it didn’t waste bandwidth and cause collisions. Imagine if two computers transmitted data at the same time: the packets of information would collide and the information would be corrupted. We would have to rebroadcast the data through the process of Ethernet Carrier Sense Multiple Access with Collision Detection- but you can just call it CSMA/CD. In simple terms, it is the protocol we use to resend data after a collision occurs.

Collisions are obviously a problem with hubs, but we just hit the tip of the iceberg. More importantly, hubs are notorious for wasting bandwidth. Hubs operate at half-duplex, which means data can only flow one way at a time. (As compared to full-duplex, where we can send and receive data between two devices at the same time.) Since we are operating at half-duplex, the bandwidth must be shared between each port on the hub. Imagine you have a 20 port hub with a 20Kb/s line to share. Tough luck, but you’ll only get 1Kb/s to each computer on the network.

What is a Switch?

A switch operates on the Data Link Layer of the OSI model. This means that switches are fairly smarter than hubs, as they can route data on a dynamic level. If information is destined for computer A, for example, the switch will only send the data to computer A.

This addresses our collision problem, since switches use what is called microsegmentation. Microsegmentation will allow for multiple collision domains. Which would you rather have- one collision domain or many?

Let’s look at an example. In the below diagram, there are many collision domains (or segments) for the switch network. If computer A and computer B sent data to each other at the same time, for instance, you may have a collision. Computer A and computer C or D, however, will not experience a collision in the process. In the hub network, there is just one collision domain. That means if the first computer wants to transmit data, it could be interrupted by any of the other computers on the network.

But the fun doesn’t stop there! Switches can keep track of which computer address belongs to a certain port. If there is information destined for computer A, it will only travel through computer A’s port. Remember how the hub split the bandwidth between each port? Microsegmentation allows for us to keep the bandwidth as high as possible for each computer. If you have a 20Kb/s line coming in, each computer can have the full 20Kb/s. (But note that switches aren’t magic, if two or more computers are using the line at the same time, they must share it. Still- this is much better than hubs, which automatically split up bandwidth to ports that you might not even be using!)

Should You Upgrade to a Switch?

The short answer here is: definitely. Hubs were primarily used because they are inexpensive and easy to setup. The good thing about “dumb” devices is that they don’t require much configuration or maintenance.

Switches are a bit more expensive, and allow for more configuration options. They can be programmed much like routers, something hubs can’t accomplish. The average home user likely won’t need the extra configuration options- but they will appreciate the speed. If your network is already blazing fast, you may not need the upgrade. If you could appreciate some extra performance, or if you have many computers on the same network, then the upgrade is almost mandatory.