What you see aboove is a Rackmounted Ethernet Switch. Every
commercial grade network has at least one. Rackmounted Ethernet
Switches are not only more powerfull than their residential
counterpart, but are completly scalable. Ethernet switches,
unlike their residential counterpart, can be daisy chaine together
in order to expand the network.
Current trend dictates incorporating as much "computer technology" (IP Phones, IP Printers, IP Cameras, Computers, Ethernet Switches, Computers Printers, etc) as possible into an existing, or erecting, infrastructure. These "nodes", as they are technically called, range everywhere from the simple embedded systems (Printers, smoke detectors, IP Cameras, etc.) to the more complex general purpose systems (Desktops, laptops, tablets, servers, etc). Although some of these are standalone nodes, nowadays most if not all are expected to intercommunicate with other nodes. The fact that the nodes are scattered throughout the infrastructures does not help the situation. All the nodes together constitute a network.
Ethernet nodes are not directly connected to each other. Instead, each and every node is directly connected to Ethernet Switch. Obviously, the more sockets a switch has, the more nodes can be integrated into the network. Furthermore, whenever two nodes want to "conversate", the source node sends the message the the Ethernet Switch and the Switch forwards the message to the destination node. This network topology, officially called start topology, although cumbersome and expensive, is the most resilient against network failures. In a start topology, should any of the hundreds of nodes ever happen to malfunction, the rest of the network is unaffected. This is contrary to ring topology were one malfunctioning node can bring down the entire network.
What makes Ethernet switches extraordinary is the fact that they dynamically create, and destroy, a dedicated connection between the two conversating nodes. Although this might not sound like much, this simple alteration drastically increase not only data throughput, but also network security. In order to understand how this is possible, it is essential to understand the alternative, however outdated, Ethernet Hub technology. Normally, new technology is suppose fully replace older technology. Astonishingly, the newer Switch technology has not, and probably never will, fully replace Hub technology. Although Hub has numerous drawback when compared to Switches, it has three main advantages which has prevented its demise. The fact that Ethernet Hubs are cheaper than Switches, compatible with Switches, and are a drop in replacement for Switches will forever prevents them from fully being replaced by Switches. The problem with Ethernet Hub based network all the nodes have whats it called a shared medium.
As its name implies, a shared medium means every node in the network "uses the same cable" (collision domain) to transmit and receive digital information. As one can conclude, this setup plagued with issues.
- Data collision - When "only one" cable is used, collision are bound to happen. As one be deduce, a collision occurs when two nodes try to transmit at the same time. Unsurprisingly, collision not only results in corrupted data, but requires the re-transmission of the data. The likelihood of a collision occurring is directly proportional to the number of nodes present in a network. Unsurprisingly, the greater the number of nodes, the greater the probability of a collision. If collisions are occurring sparingly, then they are considered a nusence. However, if collisions are occurring frequently, they can bring down the network bandwidth.
- Bandwidth Sharing - Whenever a node has to wait for the transmission cable to be free so that it can transmit, its bandwidth is inherently being divided. The severity of bandwidth sharing in a Hub network is directly proportional to both the number of nodes and how much data each node is transmitting. The greater the number of nodes in a Hub network, the more severely the bandwidth is being divided. Similarly, the more data the nodes on the network are transmiting, the greater the bandwidth is being shared. In a Hub based network, a single rogue node can bog down an entire network simply by constantly generating and constantly transmitting data.
- Network Security - The third, and probably, most serious problem with Ethernet Hubs is the lack of security. In a Hub based network, any message transmitted by any node is received by all the nodes to the network. Non-Intended recipients are suppose to voluntarily discard any packet not destined to itself. As one can conclude, this setup fraught with vonabilities. Imagine a would happen if a malicious node joined the network. Furthermore, imagine what would happen if a random node send a username and password to another node? In a Hub based network, unless the data is encrypted, the malicious node could steal the username and password.
Switches were the next generation of networking equipment which were suppose to replace the older Hub technology. Switches work by internally spawning temporary dedicated links between the two conversating nodes. A simple way of visualizing what happens internally a Switch is by pretending not only that Switches can dynamically create Ethernet cables, but use them to link two conversating nodes. As a matter of fact, what a Switch does can be mimicked simply by using an Ethernet Crossover Cable. An Ethernet Crossover Cable is used to directly link two node without the use of a Switch or Hub. The only difference from a Switch and manual Crossover Cable is that a Switch can dynamically spawn, and destroy, as many of these dynamic links as it desires. A Switch can link any node to any other node. An Ethernet Switch resolves many of the issues that are plagued by Hubs;.
- Data collision - The twisted pair Ethernet protocol standard mandates one dedicated channel for transmission and another dedicated channel for reception. What this equates to is that twisted pair Ethernet protocol is fully duplex which means both nodes can transmit at the same time. Because a Switch can dynamically spawn a temporary dedicated links between the two conversating nodes and because each link is composed of a RX and TX channel, a collision is technically unfeasible.
- Bandwidth Sharing - Whenever a node has to wait to transmit data, what is technically happening is the bandwidth is being divided. The longer the node has to wait, the more the bandwidth is being divided. In a Switch network, not only is there no need to share the bandwidth, but it is not possible to share the bandwidth. Because the Switch dynamically spawns a link between two conversating node and because each link is fully duplex, both nodes can not only transmit randomly but can transmit concurrently. Interesting fact is that a Switch does not increase the throughput but rather stops it from being divided.
- Network Security - Network security is also inherently resolved in a Switch based network as the only two nodes allowed on the temporary link are the conversating nodes. This configuration inherently eliminates middle attack. In fact, Switch based networks are so secure that, in theory, two nodes can transmit in unencrypted.