IP v6  Networking - Differences Between IPv4, IPv6, and Other Protocols

It has been said2 that the IP protocol family looks like an hourglass. The hourglass is wide at the top, where there are many application protocols, and narrows down to a much smaller set of transport protocols that are used between the two systems that take part in any particular communication session, such as TCP and UDP. A single Internet Protocol layer that is responsible for getting the packets across the underlying infrastructure forms the narrow middle of the hourglass. Below IP, the glass gets wider again to accommodate the different link layer protocols that know how to get packets from one IP router to the next, such as Ethernet, ATM, and PPP. Each datalink protocol can typically run over a variety of physical protocols that are responsible for getting the individual bits across a wire.

The hourglass model puts IP squarely in the middle of the protocol family, sitting between the low-level protocols that are different at each hop along the way on the one hand, and the high-layer protocols that function end-to-end on the other hand. This model makes IP the only part of the protocol family that must be supported on all hosts and all routers. This isn’t true for any other layer. For instance, when two hosts want to use the new SCTP protocol on top of IP rather than TCP, they can just go ahead and do so: the routers along the way don’t have to understand SCTP. Conversely, a connection between two routers may be upgraded, for instance from Ethernet to Packet over SONET (POS), without any impact to the hosts that communicate over this link through the routers in question, as the lower layer protocols are removed and reapplied every time a packet passes a router.

The job of the Internet Protocol and alternative network layer protocols (that occupy the same place in different hourglasses) is to make the packets flow from the source to their destination and to accommodate the requirements of the different lower-layer protocols encountered along the way. IP implements an “unreliable datagram service,” which means that packets (“datagrams”) can be sent from one host connected to the network to another host that is also connected to the network without first having to set up a connection. In most cases, the datagram will be delivered to the destination, but there are no guarantees. For the network to deliver these datagrams to their destination, the packet must be completely self-contained and include at the very least source and destination addresses and the higher-layer protocol to which the packet is addressed. Routers along the way look at the address to decide which way the packet should go. Routers make these decisions with the aid of the routing table, which is nothing more than a long list of destination address ranges along with pointers to neighboring routers that are willing to forward packets for these addresses into the right direction. When the destination address can’t be found in the routing table, or there is another problem, routers send back Internet Control Message Protocol (ICMP) messages to inform the source of the offending packet of the problem.

At first glance, there is significant overlap between what happens at the network layer in IP and the datalink layer in protocols such as Ethernet. Ethernet switches also use address information in the packet to forward it to the right destination. But there is a crucial difference:
Ethernet addresses are burned into the Ethernet chip, so there is no rhyme or reason to where on the planet a particular Ethernet address is used. This way of assigning addresses makes it impossible to build really big networks with just Ethernet: the routing tables would get too large. Network layer protocols, on the other hand, divide the address in two parts: the network part and the host part. Whenever hosts are connected together by using a datalink layer network, all those hosts share the same network address. The host part is different for each host (or router), of course. The network and host parts together make up the full address, which makes it possible for routers to keep track of huge numbers of hosts just by having a limited number of network addresses in their routing tables.
Having both a network layer address and a datalink layer Media Access Control (MAC) address means that there must be some kind of mechanism to map from one to the other. Different network layer protocols have implemented this in different ways and have used wildly different address lengths.

Comment an article   Name   E-mail    Title Available characters: 600 Enter what you see:

No comment posted