I’m trying to figure out a few things about network design, and I think I’m starting to get it. It used to be that you could subnet simply by changing a number in each decimal position. For example, if you wanted to have two separate subnets, you could have:

192.168.1.1-192.168.1.254

and

192.168.2.1-192.168.2.254

The two subnets would need a route to get from one to the other, they are two separate logical network sections. However, the structure of ip addressing is not the most efficient, you might only need 50 ips on one network, and 100 on the other. You can instead use classless inter domain routing to specify subnets.

Furthermore, you can isolate these ip blocks using vlans.

At least, that’s how I think routing works! Remember, most of the posts here on Docunext are simply my notes.

I think I’m starting to get it! Here’s what I’m thinking:

  • A network can have a scope, which defines how big it is, meaning how many nodes it has.
  • The scope of a network can be specified many ways:
    1. Physically, by using separate hardware and wires, duh.
    2. By using a different classful ip addressing subnet, i.e. 192.168.0.x versus 192.168.2.x
    3. By using CIDR: classless inter domain routing ip addressing, such as 192.168.0.4/27 and 192.168.0.192/27
    4. By using VLANs to separate the two subnets
  • Networks can be connected together via routers, which tell a network which IP address to use as the gateway to get to the other network.

I’m not sure if the methods of separating subnets mentioned above, CIDR and VLANs, are the same thing or not. No they definitely are NOT.

Let me take a step back with some TCP/IP basics:

  • Multihomed host: a host that's connected to two or more networks
  • An ip address is actually made up of two separate pieces: the subnet id and the host id! Like 192.168.0.1 is made up of two sections: 192.168.0 is the network subnet address, and 1 is the host address.

The most basic form of subnetting by class, then you can get more granular with cidr and subnet masks. Even with the ip address utilization efficiency gained from CIDR, you are still able to more efficiently use ip addresses more efficiently through the use of VLANs.

AWESOME! I just found a really helpful PDF from IBM. The PDF is available here: IP Network Design Guide. The pages I’m most pleased about are in Chapter 8. “Internetwork Design Study”, page 249. They go on to show good network designs for a small company, a big company, and more. This is exactly what I was looking for!

From what I’ve gathered so far, for large scale networks, the best way to connect the subnets together is with OSPF. Along these lines, a big network is made up of areas, and areas are made up of groups of ip chunks, and those individual ip chunks are subnets. Right? I think so.

With OSPF, you have a few different “sections” which are connected by special connectors: area border routers, internal router, designated router, and more. These connect the areas, of which there are several types: stubby, not-so-stubby, and totally stubby. I kid you not.

Routing related pages at Wikipedia:

Vlan

CIDR

OSPF

Other routing related page:

OSPF on OpenBSD

Autonomous Systems Overview

Online Subnet Calculator

VLAN Overview