How does it all stack up?

It seems that everything today has some level of connectivity and, subsequently, there are any number of possible ways to achieve it, mostly through employing some kind of protocol. But the embedded world is a lot different to the ordinary networked appliance, such as a server, particularly when it comes to resources such as memory. Choosing the right embedded protocol is only half the story, however, because typically one isn't going to be enough. Building up layers of protocol stacks can quickly consume even the most generous amounts of embedded memory, so where do you draw the line? The answer is, of course, application dependent but, typically, the starting point will be a TCP/IP stack. Colin Walls of Mentor Graphics' Embedded Systems Division, explained that the Internet Protocol – IP – now forms the default transfer protocol for any 'connected' device. The TCP/IP layer acts as the 'hub' for most other protocols running within an application, with other protocols either sitting above or below the TCP/IP stack. For instance, an SNMP (simple network management protocol) would sit above the TCP/IP stack (which occupies layer 3 in the OSI reference model), whilst a PPP (point to point protocol) would sit below it. While both would be 'TCP/IP aware', applications running on the device would typically only be concerned with interfacing at the TCP/IP level and may or may not be aware that its requests are carried out by either SNMP or PPP. So if TCP/IP is the all seeing eye that manages connectivity, are the other protocols really necessary?