In these articles, I would like to present how to build a basic MPLS network.  Whether you just wondered how your service providers network was built or if your considering an internal MPLS network inside your company, this information should help out.  As far as why you would want to build an MPLS network, I won’t go into much detail as this is more intended to show how to build and not why to build.  However, I will say the big reason to build an internal MPLS network is so that you can use MPLS based VPN’s using Virtual Route Forwarders or VRF.  That’s all I’ll mention on that, lets starting building a network.

I’ll be working with Cisco 7200 routers, since dynamips is being used as the platform.  We could just as easily use some of the other emulated hardware, but out of all the options available in dynamips – a 7200 is probably the closest to what you might see in the core of an MPLS network.  There are three terms that describe the type of MPLS device:

P – which is for Provider equipment
PE – which stands for Provider Edge
CE – which as you have probably already guessed, stands for customer edge.

As far as a provider network that you use for a company WAN solution, your equipment is obviously the CE device and the PE device is what directly connects you into the providers network.  The P devices is what you would consider the providers backbone and would never ‘see’ or interface with that hardware.

In an internally built MPLS network, the premise (or remote office) routers would still be considered the CE equipment and the PE device would typically be in a company data-center serving in a traditional ‘distribution layer’.  The P devices in a corporate built MPLS network would be what connects multiple data-centers together.  For example, if you have two data-centers, then you would have at least one P device at each one that provides the connectivity between those two sites.  The type of hardware used for the P device would be something that supports MPLS label switching.  This could be 7200 routers, 7600 routers, a 6500 Multilayer Switch, or others.  There would be a PE device directly connected to the P devices at each of the two data-centers.  The PE device would be what represents your typical distribution layer, which could be a LAN distribution switch, or a WAN router, or both.  The CE equipment would be say a router at a remote branch office, which resides on the edge of a WAN.  The CE device could also be a firewall back in the data-center.  As you can see, there are many different configurations that one could encounter based on the needs of the company.  I have just tried to list a few of the more common scenarios, but by far is not an exhaustive list.

A little about the lab

As I mentioned earlier, we will be using Dynamips to provide the platform on which we will lab this exercise.  More specifically – GNS3, the all encompasing wrapper around dynamips (and now many other programs I might add) will be the tool of choice here.  At the end of this article, you will find an importable project that you can load directly into your own GNS3 installation.  However, to get the very most out of the exercise – I would recommend building the network out manually in GNS3, to get the complete feeling of the build.

CompanyX network scenario

CompanyX has two datacenters, that are connected by high speed point to point connections.  We have two 100Mbit connections between the datacenters, each one provided by a different commercial carrier.  Each datacenter has two ‘core’ devices which will terminate a single 100Mbit connection.  The two core devices in each datacenter will have a local connection to each other, so that full connectivity can be maintained in the event one of the two metro connections are lost.  These ‘core’ devices will serve as the ‘P’ devices, which does nothing but what is called ‘label switching’.  By the time user traffic reaches these devices, they have labels appended to the packet headers and that is what is used to determine where to foward that traffic.  The concept here applies just like when Layer3 switching was introduced – it is much quicker to switch a packet than it is to route it.

CompanyX network diagram containing P devices

The point to point connections are all /30 networks and using the 10.1.1.0 network. All loopback addresses are assigned out of the 10.254.1.0 network as /32 or host addresses.

So, let’s dive in… Below you will find the basic GNS3 project information that is importable – otherwise use the diagram above to build out the network.

Here is the relevant configuration for each router:

hostname r1
!
interface Loopback1
 ip address 10.254.1.1 255.255.255.255
!
interface FastEthernet1/0
 description r2 f1/0
 ip address 10.1.1.1 255.255.255.252
!
interface FastEthernet1/1
 description r3 f1/1
 ip address 10.1.1.5 255.255.255.252

hostname r2
!
interface Loopback1
 ip address 10.254.1.2 255.255.255.255
!
interface FastEthernet1/0
 description r1 f1/0
 ip address 10.1.1.2 255.255.255.252
!
interface FastEthernet1/1
 description r4 f1/1
 ip address 10.1.1.13 255.255.255.252

hostname r3
!
interface Loopback1
 ip address 10.254.1.3 255.255.255.255
!
interface FastEthernet1/0
 description r4 f1/0
 ip address 10.1.1.9 255.255.255.252
!
interface FastEthernet1/1
 description r1 f1/1
 ip address 10.1.1.6 255.255.255.252

hostname r4
!
interface Loopback1
 ip address 10.254.1.4 255.255.255.255
!
interface FastEthernet1/0
 description r3 f1/0
 ip address 10.1.1.10 255.255.255.252
!
interface FastEthernet1/1
 description r2 f1/1
 ip address 10.1.1.14 255.255.255.252

GNS3 project for basic MPLS build