What Makes a Switch?
Switches can vary greatly in the types of materials used in their construction, ranging from metal to plastic for the housing. Plastic tends to be the material of choice among lower-end models. However, companies like Netgear offer unmanaged products, such as its GS105, with metal housings. They last longer than plastic and tend to be more durable.
Smaller unmanaged switches tend to draw their power from wall warts. Higher-end models (like those designed to be rack mounted) usually have an in-housing power supply and only require a standard IEC power cable. As with any electronic device, switches will generate heat that must be dissipated. Many smaller switches will be passively cooled utilizing a heat sink (such as pictured above). However, larger models are normally equipped with fans to help exhaust the generated heat and prolong the life of the product.
A Broadcom SoC inside the Netgear GS608 8-port 10/100/1000 Mb/s switch
Companies like Broadcom and Qualcomm supply a lot of the component hardware used in consumer networking products like switches, as well as other product lines ranging from routers to modems. Using this hardware, the switch knows how to handle the incoming and outgoing information. How these instruction are dealt with varies between unmanaged, smart and managed switches. For example, many unmanaged switches utilize SoC (or system on a chip) technology, where many subsystems are integrated on the same package. This SoC handles the data at the hardware level when it comes into the switch, and directs it where to go. For example, in a switch’s case, a frame would come in and be compared against a MAC table, then be directed out the appropriate network port. Higher-level switches, such as smart switches, generally contain CPUs that are integrated into a chip within the switch. This CPU can aid in handling the additional overhead of features like VLANs, QoS and Layer 3 switching.
A convenient feature added to switches is the inclusion of link lights, which are generally located near the top of the port and on either side of the port as well. The lights activate when the port detects activity from an active connection, which occurs when two products are linked together through the use of patch cables. For the most part, one activated light means that the established connection is 10/100 Mb/s, while two activated lights indicate a 1 Gb/s connection. This can be extremely useful for troubleshooting purposes. Even though link lights do not guarantee a reliable connection, they at least indicate that two devices are patched through to each other. If there are no lights, it can create a starting point for troubleshooting in that the port could potentially be bad or that the cabling between the two connected points is defective in some manner.
Apart from the physical aspects that make up a switch, there are also different features to look out for, such as jumbo frame support, duplexing, large MAC table sizes and Auto-MDIX/MDI support.
Jumbo frames are Ethernet frames that contain more than 1500 bytes, with 9000 bytes tending to be used as a conventional MTU size. Jumbo frame support can be useful for lowering bandwidth requirements, as the switch’s CPU only has to process one larger frame as opposed to being flooded with multiple smaller frames.
There are two types of duplexing you'll encounter: full and half. Most modern equipment operates in full-duplex mode. This is very important, as a device running full duplex mode is able to simultaneously send and receive data. In half-duplex mode, one device sends a frame and all devices within the collision domain listen for the frame. This has the potential to introduce collisions if collision detection fails to prevent the occurrence, which can significantly degrade the performance of your network.
Large MAC table sizes are more of a business-oriented capability than something that'd come into play at home. Businesses need a switch with the capacity to retain a large MAC table, as there are most likely hundreds of device IDs that will be stored in the switch’s table. This is less problematic when the number of clients on your network is small. However, it is still a desirable feature for preventing issues that could arise from running out of space on the table. At that point, the switch would not be able to properly route information, which would significantly impact performance.
Auto-MDIX/MDI support is becoming commonplace. Without this feature, devices that contain MDI (media independent interface) or MDIX (media independent crossover) connections on both sides, such as switch-to-switch connections or peer-to-peer computer connections, would require a cross-over cable as opposed to a straight-through. Auto-MDIX/MDI allows devices to be connected to each other using either type of cable, as Auto-MDIX automatically detects the required cable type and configures the connection appropriately. This removes a great amount of confusion while interconnecting devices, as two types of cable connections are no longer needed.