THE FOUR-LAYERED TCP/IP STANDARD
The seven OSI layers don’t directly mirror any particular networking system, and that’s arguably a strength. It means the model can be used to analyse or plan out any sort of networking scenario.
However, not everyone agrees that OSI is a good fit for the 21st century. That’s because the vast majority of modern networks use TCP/IP – and this particular collection of standards is defined in terms of four layers, which don’t perfectly match the OSI divisions.
In TCP/IP, the bottom layer is the link layer, or network access layer. This includes defined standards for both connection protocols (such as Ethernet and Wi-Fi) and basic communications over those connections, so it’s like a combination of OSI Layers 1 and 2.
The next two TCP/IP layers are the internet layer and the transport layer. These cover much the same ground as OSI Layers 3 and 4 respectively. The name internet layer isn’t exclusively for internet connections, but refers to the general ability to route packets between different LANs – something the OSI model includes in its more loosely named network layer.
The major divergence from the OSI model is that TCP/IP doesn’t recognise session and presentation management as separate layers. Its fourth and top layer is the application layer, which may optionally include software routines that handle encryption, character encoding and other such matters.
With only four layers, the TCP/IP model is simpler to work with. And because the layers relate directly to published specifications and standards, it is more practical for troubleshooting and network architecting.
Even so, the OSI model is almost 50 years old, and its broad applicability means it’s not going anywhere soon. When network engineers talk about “Layer 2 load balancing” or “Layer 3 switch hardware”, they’re using the OSI definitions. They’re universal terms that are understood across networks of all types, and will probably still be long after TCP/IP has fallen by the wayside.