1, latency

The delay is defined as the time at which the switch begins to receive information frames until it starts forwarding frames. The length of the delay depends mainly on the technology used by the switch. Such as a fast forwarding technology switch, since it only begins to forward the physical address of the message frame, so the delay it generates is fixed, typically 40 microseconds. The delay generated by the switch that stores the forwarding technology is related to the size of the information frame. Because the switch to read all the information frame to the buffer and then forward. So a 1000-byte message frame running on 10Mbps Ethernet has a delay of 800 microseconds.

2, Single & Multi MAC

A single physical address switch associates the address of a hardware with a specific switch port. The main purpose of its design is to add a single user or a shared resource to a switch port. The multi-physical address switch allows multiple hardware addresses to be connected to one port of the switch. So the switch can be connected to a mouth of a media shared hub, which can support multiple end users. The number of addresses that the switch can support is different for each switch vendor. If the number of hardware addresses exceeds the storage capacity of the forwarding database provided by the switch, the switch may send the information frame to all ports with a new address, or even throw the information frame completely.

3, Spanning Tree

General Ethernet does not allow more than two devices to connect two network segments together. There are two or more paths between segments or subnet that can produce BridgeLoop.

Because the basic working state of the switch is similar to that of a multi-port bridge, they produce the same problem as the bridge structure network, including the bridge ring. That is, a broadcast from one network segment through the bridge to another network segment and then through another bridge (path) back to send the broadcast network segment. The spanning tree protocol (typically IEEE802.1d or DEC) makes the devices with bridge functions in the network communicate with each other, using the spanning tree algorithm to select the best transmission path, and the other paths will be blocked. In addition, the spanning tree algorithm also monitors the success or failure of the path. If it fails, it will automatically create an alternative path. The bridge ring may be caused by an unexpected connection, but it may also be intentionally inserted, and its purpose is to provide an alternate path. If the switch does not support spanning tree algorithms, it is important to avoid creating bridge loops.

4, management

For the network managers and the face of different manufacturers to provide different switches, the switch management is a problem. There are two basic methods to choose from. The first method is to add the management function on the backplane, it will send the information frame to the statistics and store it in the management structure. This method can access the switch with SNMP anywhere in the network and report statistics based on MIBII or RM0N standards. The second method is the refraction or "mirror" method. At this time the switch will be seen on one or more of the mouth of the forwarding traffic to a designated port, called "monitoring mouth." Unfortunately, this method does not have a uniform standard, each vendor is in accordance with their own specific methods to achieve.

5, full duplex

Full-duplex Ethernet performance is much better than switching Ethernet. Full-duplex requires a cable connection that allows the switch and the site to transmit and receive at the same time. This means that full duplex at this time can achieve 100% utilization. In theory, there is a 200% utilization in the absence of a collision. But duplex requires a lot of investment. Because the switch requires full duplex, the network equipment on the network card also be able to support full-duplex. This is not practical for most web users.