What Is an IP Address?

Let’s look now at how IP packets are addressed and how a packet finds its destination. We will do this through a series of analogies, eventually coming back to our postal service analogy.

The question we’re going to answer is: How does a packet find its destination?

The internet consists of millions of routers and pipes that interconnect billions of hosts. A host is simply an endpoint that communicates over the internet. Hosts include servers, laptops, smartphones, and other devices. When a host wants to send a packet to another host, that packet has to traverse a path through the internet, specifically through the routers and pipes that make up the internet. The pipes are the physical links that connect routers to each other and to hosts.

Routers connect multiple pipes and have to make the routing decisions. That is, when a packet comes in on a link, the router has to determine which link to send it out on. This is the job of the router’s routing table. To understand routing tables, we will go through a series of analogies.

For our first analogy, let’s look at the internet as a network of streets. In this case, a packet going from host A to host B is like driving from city A to city B. A router, then, is like an intersection where multiple streets come together. When driving and we get to an intersection, we have to determine which way to turn, that is, which street to turn on to. A routing table, then, is like the direction sign that we hope to see at the intersection.

This analogy breaks, because a street sign cannot possibly list every possible city to which we might be going. Analogously, a routing table cannot possibly list every possible host to which a packet might be destined. In other words, a packet’s destination address cannot be something like “Apple web server.”

An analogy that does work, though, is rooms on a hotel floor. In this case, a packet going from host A to host B is like walking from room A to room B. A router, then, is like an intersection where multiple corridors meet up, and the routing table is like the sign at the intersection that says rooms 201–220 to the right and rooms 221–240 to the left. If I’m going to room 211, I know that I turn right.

Why does this work? Well, for one thing, all of the room numbers in a range are in the same place. So we don’t need directions to each room, but just to each range. Also, importantly, all possible room numbers are covered by the ranges on the sign. There are no rooms numbered outside the range 201–240.

We can do the same thing on the internet. So rather than addressing hosts with names, we’re going to use numbers.

These numbers are called IP addresses, which are (32-bit) numbers. They are usually written in dotted notation, for example, 73.227.130.85. The numbers between the dots are called octets (because they are 8 bits). Hosts are located in contiguous ranges of IP addresses. For example, all hosts with addresses that start with 128.30.*. * and 128.31.*. * are at the Massachusetts Institute of Technology (MIT). Routers only need directions for each range. These ranges are called prefixes or CIDRs, which stands for Classless Inter-Domain Routing.

Let’s return now to our postal service analogy. Now, a packet going from host A to host B is like a letter sent from home A to home B. A router is like a post office. Letters come in on trucks and then have to be sorted to go out on other trucks, so a routing table is like the sorting rules for the letters. So let’s look more carefully at how letters are addressed and sorted.

Continuing our analogy, a packet addressed to 128.30.27.149 is analogous to a letter addressed to Lucille Ball, 1000 Roxbury Drive, Beverly Hills, CA 90210. To route this letter, most post offices don’t have to look at the whole address. In fact, only the 90210 post office has to look at the street address. All other post offices only have to look at the zip code. That is, the letter can go all the way from the sender to the 90210 post office, based only on the zip code.

We do the same thing with IP addresses. Each address is split into a left side and right side. The left side is called the prefix, and it is analogous to the zip code. Recall that all IP addresses with prefix 128.30 belong to MIT. So only the routers in the MIT network have to look at the whole IP address. All other routers on the internet only have to look at the prefix. That is, the packet can go all the way from the sender to the MIT network based only on the prefix.

A quick note: The divide between prefix and the rest of the IP address does not have to occur in the middle or even at an octet boundary, though for simplicity in this presentation, all examples will put the divide in the middle so the prefix is the first two octets.

Now we can see that instead of addressing IP packets with names, we have to address them with numbers, that is, with IP addresses. So instead of something like “Apple web server,” we have something like 17.253.207.54. And instead of “Bobby’s laptop,” we have 73.227.130.85.

We can also see now how a routing table works. A routing table simply maps prefixes to links, and the router sends the packet out on the link corresponding to the prefix that matches the destination IP address. In this example, the destination IP address is 17.253.207.54, which matches the prefix 17.253 on the second row of the table, so the packet goes out on link C.

The key point is that rather than having to list every possible host name or even every possible IP address, routing tables only need to list every possible prefix. Remember it’s analogous to routing letters by zip code instead of having to consider every possible street address.

There are four types of IP addresses: public, private, static, and dynamic.

The divide between prefix and the rest of the IP address does not have to occur in the middle or even at an octet boundary.