Ethernet over MPLS

Multiprotocol Label Switching (MPLS) is used by service providers to implement QoS, tag switching, service levels, and many other features. The service is very popular, especially in Europe and Asia. Ethernet over MPLS (EoMPLS) is a Cisco solution that is currently in an RFC draft state with IETF. It extends MPLS by tunneling Layer 2 Ethernet frames across a service provider's Layer 3 core. Doing so provides two advantages:

  • The service provider has more scalability because it has a Layer 3 core.

  • Your Layer 2 information, including STP, can be tunneled through the service provider.

Because of these two advantages, service providers prefer EoMPLS over Q-in-Q. EoMPLS, like Q-in-Q, is a tunneling mechanism that tunnels your VLAN information across a service provider's network. The main advantage that EoMPLS has over Q-in-Q is that EoMPLS supports more than 4,096 VLANs by the service provider.

graphics/alert_icon.gif

EoMPLS extends MPLS by tunneling Layer 2 Ethernet frames across a service provider's Layer 3 core. EoMPLS has more scalability because it has a Layer 3 core and Layer 2 information, including STP, can be tunneled through the service provider. EoMPLS scales better than Q-in-Q.


Overview

EoMPLS can deliver Transport Layer Security (TLS) for customers' Ethernet connections. TLS provides a logical connection between two sites across a point-to-point connection. From the customer's perspective, this logical connection appears as an Ethernet segment. Some of the advantages that EoMPLS have are that because EoMPLS is based on a Layer 3 process, Layer 2 problems and management are not an issue for the service provider. For instance, with Q-in-Q, which is a Layer 2 process, the provider must deal with internal STP, MAC address learning and forwarding, and other Layer 2 processes. EoMPLS with TLS doesn't have this limitation because the provider deals with internal traffic from a Layer 3 perspective. This provides much more scalability and control over traffic.

Process

Before I begin discussing how EoMPLS functions, you need to be familiar with some important terms that MPLS uses, as shown in Table 11.2.

Table 11.2. MPLS Terms

Term

Definition

Label distribution protocol (LDP)

LDP is a protocol that defines labels that are used to classify traffic and how the traffic should be treated inside the network.

Label switch router (LSR)

The LSR switches labeled frames between interfaces and can be either a router or a switch (like an ATM switch).

Edge label switch router (Edge LSR or LER)

The LER takes traffic from the customer, labels it, and switches the labeled frames; it is also responsible for stripping off labels on egress ports.

Label switch controller (LSC)

The LSC is typically a router that controls an ATM switch; in other words, it is an MLS-based ATM switch that perform Layer 3 switching.

graphics/alert_icon.gif

Remember the terms in Table 11.2.


Figure 11.8 displays a sample provider network. In this example, an LER takes ingress traffic from the customer and labels it. Cisco 7600 routers can function as an LER. LSRs inside the network use these labels to perform switching. LDP determines how the service provider's gear will treat and process the labeled frame. An LSC is an MLS-based switch that can perform switching of Layer 3 information at Layer 2 speeds. It is typically a hybrid router/ATM switch.

Figure 11.8. Service provider and EoMPLS.

graphics/11fig08.gif

Protocol Labeling

EoMPLS is implemented by a service provider and is a point-to-point connection, with LERs being the endpoints of the connection. The ingress LER attaches two labels to incoming frames: a tunnel and a virtual circuit (VC) label. The tunnel label is used to determine what egress LER device the traffic should be forwarded to. The VC label determines the egress port on the egress device.

It is important to point out that each customer needs its own physical interface on an LER. Each customer typically has one VC associated with the interface. However, if more than one VC is associated with the interface, the customer must tell the serviced provider how traffic should be mapped to specific VCs.

The ingress LER performs two functions on a frame received from the customer: frame marking/classification and encapsulation. On receiving a frame, the LER first maps the frame to a tunnel label switch path (LSP), which is the path that the frame will take through the provider's network. Next, the LER marks the frame with a CoS value, which becomes part of the tunnel tag. With DiffServ, the frame is marked either E-LSP (queuing, scheduling, and drop policy information) or L-LSP (drop policy information). The CoS information is inserted into a tunnel label in a 3-bit field called EXP. The CoS can be statically assigned by the provider based on how the customer purchased the service, or the provider can map the 802.1Q/P information from the customer's frame into the equivalent CoS that the provider has configured.

The ingress LER then adds the VC label, which is used by the egress LER to forward the traffic out the correct destination port. Both the tunnel and VC labels are included in an EoMPLS encapsulation, as shown in Figure 11.9.

Figure 11.9. EoMPLS encapsulation.

graphics/11fig09.gif

When an internal LSR receives the labeled frame, it examines the destination MAC address to determine whether it needs to process the frame. The LSR then examines the tunnel tag to determine how to switch the frame. When switching the frame, it rewrites the Layer 2 header according to its own source and next-hop's destination MAC addresses.

When the egress LER receives the labeled frame, it removes the header and tunnel label. The LER examines the VC label to determine which physical interface the frame should exit, and then strips this off and queues up the frame on the egress port.

graphics/alert_icon.gif

EoMPLS uses two tags: a tunnel tag and a VC tag. The tunnel tag describes how to get the user's data across the EoMPLS network, and contains CoS information. The VC tag is used by the egress carrier device to determine the exit port to use to forward the frame to the customer.


Connection Types

EoMPLS currently offers point-to-point connections. Development on point-to-multipoint is being worked on. The next two sections examine these two types of connection solutions.

Point-to-Point

Providers like point-to-point solutions because they're easy to provision and maintain, and are compatible with a backbone solution that uses MPLS. With EoMPLS, you have better service provider scalability than with Q-in-Q because you aren't limited to 4,096 VLANs in the provider's core. The provider can actually use up to 20 bits to differentiate between customers, even in a fully meshed network.

However, point-to-point connections have problems fully meshing a network because it cannot be done via trunking. You have to use separate VLANs for separate sites for connectivity, where the provider separates the traffic across different VCs. You then need an RP to route between the VLANs. You could use separate physical connections between different sites, but this would increase your costs. Either way, there are customer scalability problems with point-to-point connections.

Multipoint

In a multipoint EoMPLS solution, the service provider emulates an Ethernet switch. This is typically done via a point-to-multipoint VC, which emulates a broadcast medium. From a provider's perspective, the main problem with this approach is that it is difficult to set up and maintain especially with QoS support. When many sites must be meshed, customers like this type of solution because it simplifies their connection process and the operation of their switches across the MAN.



BCMSN Exam Cram 2 (Exam Cram 642-811)
CCNP BCMSN Exam Cram 2 (Exam Cram 642-811)
ISBN: 0789729911
EAN: 2147483647
Year: 2003
Pages: 171
Authors: Richard Deal

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net