Ethernet uses source and destination access points to keep track of multiple channels of communication over the network connection. Ethernet service access points are used to map the IP communication to the physical layer. The source and destination access points are used to allow TCP/IP to use the same physical connection to support multiple logical connections. Unlike TCP/IP which has end-to-end pairings of source and destination ports, service access points have meaning only within the LAN (one subnet). The service access points are selected by the server
To use TCP/IP as an example, when a server opens a connection to a website, the HTML file is downloaded and within it are references to images, sound files and other parts of the page which are stored at the server in different files. While the web browser will use different virtual port numbers to open the connection to the server, the computer still needs to know which of the ethernet frames are meant for the pictures, which are meant for the sound file and which is meant for the java file. To keep all the connections straight