I am having some trouble understanding why does a WAN link need encapsulation, such as MPLS or frame relay? Is it merely to encrypt the info from one LAN to another LAN across the internet?
The simplest WAN topology is a point-to-point link known as leased line or private wire. No encapsulation is necessary (although, in practice, there is usually a small header and checksum, such as HDLC).
In this scenario, the telco transports your bitstream from end to end, reserving 100% of the contracted bandwith at all points on the network.
Suppose you have a head office and 20 branches in a star of point-to-point links. You will need 20 links and 20 router ports. This is expensive. If you want to create a backup data centre, you multiply the cost by two.
To save money and router ports, the telco can offer you a switched WAN.
At the head office site, you have one high-bandwith link, which encapsulates the data streams to the branches in a protocol such as frame relay or ATM (the first non-proprietary protocol of this type was X.25). Encapsulation is necessary to address the WAN endpoint. One can also give information about priority and congestion. The telco can sell you different levels of garanteed bandwith (less than the circuit capacity), which saves you money. The branches have a similar setup. If necessary, the branches can communicate directly, without passing through head office.
MPLS is a VPN overlay on these layer 2 protocols, which enables the telco to route your packets by IP address, making it much simpler to set up large and complex networks, while still having a secure private network. It is simpler (and therefore cheaper) for the telco to mix private MPLS networks and public internet on their network. They can also provide internet and private MPLS services on the same router at your site.
In all these switched WAN scenarios, adding a backup data centre requires just one extra connection, one router port and the appropriate configuration.
I should also mention Internet VPN. In this case, you are encapsulating and encrypting your data over the public network. Encapsulation allows you to access private IP addresses in the corporate network, encryption is necessary for security. What you do not have (with certain exceptions) is guaranteed bandwidth or high service levels, but it's much cheaper.
IP packets are encapsulated in a layer 2 protocol, whether they are sent on a LAN or WAN. Frame relay is one such WAN layer 2 protocol. There are others, but most are being rapidly replaced by Ethernet. On LANs, of course, you have Ethernet and Wi-fi.
MPLS doesn't quite fit into the TCP/IP or OSI model. Some consider it a "layer 2.5" protocol.
The term encapsulation is used to describe a process of adding headers and trailers around some data. For example, here is what happens when you send an email using your favourite email program (like Outlook or Thunderbird):
The email is sent from the Application layer to the Transport layer.
The Transport layer encapsulates the data and adds its own header (with its own information, such as which port will be used) and passes the data to the Internet layer
The Internet layer encapsulates the received data and adds its own header, usually with information about the source and destination IP addresses. The Internet layer than passes the data to the Network Access layer
The Network Access layer is the only layer that adds both a header and a trailer. The data is then sent through a physical network link.
Fore more details wiki
Layer 2 Encapsulation covers a number of purposes, most notably framing, error detection, compression, addressing, and protocol layering, and these really apply whether the networking is wide area or local area.
- Framing: We need some way to reliably chunk the data into frames
- Error detection: If the encapsulation has addressing or length fields, it's normally vastly much more efficient if it can do its own error detection (and perhaps correction), so it can discard frames which have been corrupted. It's good to put this at a low level, because it's much more likely to have knowledge of what kinds of errors are probably and improbable.
- Compression: it might be that there is a lot of redundant information in the frames, such as for example perhaps all the packets have pretty much the same header, and a low-level compression might be practical.
- Addressing: Very frequently there are multiple devices on the medium, and we want to be able to select which to disturb with this frame: most notable for broadcast media such as ethernet, and often not present in encapsulation for point-to-point networks.
- Protocol layering: it's very common that the technology should be capable of being used by several upper protocols, perhaps internet protocol and some other. The encapsulation has some method of saying which upper-layer protocol stack should receive it.
The simplest encapsulation is SLIP, which just does framing and nothing else: it's for point to point links, so addressing isn't used, and the only upper protocol suppported is IP, so it doesn't even have to say that. Van Jacobson's method of compressing TCP/IP headers is an example of low-level compression.
One of the most complex is ethernet, which does framing, error checking, broadcast/unicast, local addressing and upper protocol selection.