OSI Layers explained

Question

I'm trying to understand Network layers. There are five layers, application, transport, network, link and physical layer. I have studied from various sources but I couldn't understand clearly, especially transport, network and link layers. Can you explain those layers for a dummy person like me? Thanks.

Answer 1

So I think the hardest part people have with the OSI model is that they look at some of the layers and forget its talking about communication... application layer specifically people get confused as to what its talking about.

Application references the protocols applications use to communicate. Examples being ftp, http, etc.

As stated above, the original OSI model (which is really just theory, nothing actually implements it exactly as it would be too inefficient) has 2 more layers in between application and transport (which are most likely wrapped into the application layer). These are Presentation and Session.

Presentation is responsible for things like HTTP's Accept-Encoding: gzip, deflate. encryption and character encoding are said to be Presentation layer, so SSL/TLS falls under this category (the s in https).

HTTP Sessions is literally would be a literal example of the Session layer (anything that keep your application state alive across connections). If a protocol is connection oriented, it might not have a session layer, which is why HTTP is probably the only example I can think of at the moment.

As you can see HTTP was the answer to all 3 of these top layers, which is why they have all been combined into Application layer in newer versions of the model. Cisco still uses the 7 layer model, but Microsoft uses the 5 layer you're using.

Transport is TCP... it contains information about reordering packets and can adjust the amount of packets per window, allows the computer to know if anything was lost and ask for re-transmission, etc. UDP is also an example of the transport layer, however its a lot simpler of a protocol, no re-transmission of packets are done on UDP.

Network is the IP protocol (also IPX/SPX from the old netware days, and ICMP (pings) and IGMP (routers)) this allows for addressing computers that rely outside the a collision domain (things separated by switches or routers).

Link or DataLink layer is ethernet, (or ATM, or FDDI) which addresses computers physically connected to a hub or directly with a network cable to each other. This layer adds the MAC addressing in ethernet and the frame part of a packet is the header it uses.

Physical layer (in original OSI model) is just your cables and network equipment.

Pretty much the only people who still talk about OSI is network techs. They may still say Layer 2 or Layer 3 switches, etc which is in reference to OSI.. Layer 2 is an ethernet switch, a layer 3 switch adds routing.

The best way to see how this is used is to load up wireshark and snoop your own network traffic. It will actually show you the parts of a packet that are responsible for most of the layers.

Knowing the OSI model is not really overly useful, but it can help you organize the stages of network communication in your mind and help you troubleshoot.

Knowing the protocols and how they interact is extremely useful, learn how tcp negotiates connections, IP addressing and subnet masking, HTTP and Ethernet can help you whether you are a developer or a server admin, or network admin or even a DBA. There is nothing like a bad network card to ruin your weekend or even a whole month if no one thinks to check it, and the only way to tell is use tcpdump, and wireshark to see errors in the Ethernet frames.