Metro Ethernet has recently been the “buzz” of the telecom industry.  Emerging out of this is a technology that will expand its penetration to beyond the metro rings of cities into the backbone as well. This is Virtual Private LAN Service (VPLS) (news – alert). As multiple protocol options are being worked, it is important to understand their differences in design and purpose.

Early entrants to this industry have had difficulty getting out of the starting gate.  Cogent Communications, Yipes Enterprise Services (news – alert) and XO Communications (news – alert) have all met tough economic conditions.  As the market improves in 2004, the time has come for Public Ethernet Services and technologies like Virtual Private LAN Service (VPLS) to shine.   Ethernet carrier revenue is expected to grow at annual rates of more then 50 percent and exceed $4 billion by 2005, as reported by Gartner Dataquest. Services such as VPLS will start to emerge in the coming year.   VPLS offers a multipoint form of a Layer 2 Virtual Private Network (VPN).  

VPNs provide a private virtual network over a public network.  They are now a hot commodity for frame-relay replacement. MultiProtocol Label Switching (MPLS)-based IP VPNs are also becoming quite popular.  Since they leverage existing devices and offer multiple services.  IP VPNs are offered in two major types: Layer 2(L2) VPNs and Layer 3(L3) VPNs.  The Internet Engineering Task Force (IETF) (news – update) has been working on standardizing the two major types within the l2VPN and l3VPN working groups, respectively. 

The l3VPN working group is tasked with standardizing three types of Layer 3 VPNs; 2547bis, Virtual Routers, and CPE-IPSec-based.  BGP/MPLS IP VPNs (draft rfc2547bis) define a method that allows service providers to use their IP backbone to provide VPN services to their customers.  Draft Muthukrishnan defines a method that allows the creation of VPNs via virtual routers as opposed to virtual routing function (VRFs).  Lastly, the task of standardizing CE-based VPN using IPSec has fallen on the l3VPN working group.   Virtual routers, which had early success with vendors such as CoSine, Shasta, and Springtide, have met with resistance in North America.  CE-based VPNs using IPSec and 2547bis-based VPN have become the dominate method for VPN implementation.   

The l2VPN working group is tasked with standardizing Virtual Private Wire Service (VPWS), Virtual Private Local Area Network Services (VPLS) and Internet Protocol LAN Service (IPLS).   VPWS is a L2 service that provides L2 point-to-point connectivity (e.g. Frame Relay DLCI, ATM VPI/VCI, point-to-point Ethernet) across an MPLS-enabled IP network.  

The two main competing drafts for VPWS are Draft Kompella (K. Kompella) and Draft Martini.  Both drafts use the encapsulation method as defined by PWE3 (Pseudo Wire Emulation Edge to Edge) Working Group known in the industry as Martini encapsulation (named after the author of the draft).  The main difference in the two drafts is the technique for signaling the psuedowires: Draft Kompella uses BGP and Draft Martini uses LDP.  VPLS is a L2 service that provides L2 Multipoint Ethernet connectivity across an MPLS-enabled IP network. VPLS appears in all respects as a LAN to customers of a service provider.   However, in a VPLS environment the customers are not all connected to a single LAN, therefore they can be spread across a metro or wide area.  In essence, a VPLS glues several individual LANs across a packet-switched network to appear and function as a single LAN.  Lastly, IPLS is a solution for a specific topology where MAC learning capabilities are not required for VPLS service.  In this specific topology, IP host or IP routers are in place instead of LAN switches, as is the case with VPLS.

Virtual Private LAN Service, also known as Transparent LAN Service (TLS) and Virtual Private Switched Network Service (VPSNS), serve to extend an enterprise’s network over a service provider’s wide area network in a layer 2 fashion.  The two competing drafts, which are currently both hotly debated within the standards committees, are the Lasserre-V Kompella draft and the K.Kompella draft. 

Before looking at the debated differences between the two drafts, the functional components of VPLS and how they work will be addressed:

CE- This is the Customer Edge device. The CE devices that belong to a VPLS instance interact through the SP (service provider) network as if they were contacted by a LAN.  Most often this is a LAN switch, which can send either tagged or untagged Ethernet traffic.

• PE- This is the Provider Edge device.  The PE device has many duties that include:

• Martini (PWE3-Ethernet) encapsulation

• LAN switching functions like MAC learning, flooding, aging and switching on a per VPLS instance basis.

• PE Discovery – PEs discovering other PEs that are apart of the VPLS.  This function is a point of debate and contention between the drafts. 

• SP Cloud - This is the Service Provider Cloud.   The SP Cloud is responsible for control plane signaling using Label Distribution Protocol (LDP) or Border Gateway Protocol (BGP). 

The above diagram shows an example of a customer of a service provider with three geographically distinct sites.  The service provider’s IP/MPLS cloud has already been provisioned with a full mesh of LSPs for each site. The PE learns of remote hosts by associating source MAC addresses of packets with the ports on which they arrive. When the PE receives a packet to a destination that is not in its Forwarding Information Bases (FIB), it will flood the unknown destination packet to all other PEs in the VPLS.  Upon receiving the frame, the other PEs further flood the frame to the CEs, if the PE did not know the destination MAC.  VPLS uses split horizon flooding to prevent loops from forming due to the full mesh for LSPs. Once the unknown destination MAC is learned it will be a part of the FIB for that VPLS instance.

To understand the debate, the between the two drafts, the differences in concept must first be understood.  Both drafts utilize Draft Martini (PWE3-Ethernet) for the encapsulation for the data plane.  The mechanism for Pseudowire (refers to a tunnel) signaling, which performs the setup and distribution of labels, is a point of great contention.   Draft K. Kompella advocates the use of BGP and Draft Lasserre-V Kompella advocates using LDP to accomplish pseudowire signaling. The argument against using BGP mainly centers on the complexity of the BGP protocol.  On the other hand, the arguments against LDP are that LDP does not have auto-discovery.  Draft Lasserre-V Kompella does not specify any mechanism for auto-discovery.  Proponents of Draft Lasserre-V Kompella say that a centralized approach is better such as using DNS, Radius, or even Directory- based is more effective.

While the conflicting drafts may appear to represent a roadblock to a viable solution, the resulting contention may actually present the path to an optimal solution. Quite often in the telecommunications industry, drafts are adopted and finalized without recognition of and significant debate on all of the relevant issues. This results frequently from political issues that sometimes supersede the more relevant technological points of contention. For example, a more powerful and influential vendor may drive a solution that is not necessarily technically-superior to that of a vendor of lesser influence. Consequently, a final IETF draft does not always represent the most optimal technological solution. Therefore, the healthy technical debate reflected by the competing drafts bodes well in terms of arriving at the optimal technical solution. From the point of view of the service provider, they are locked into supporting the technology of their existing multi service routers or, as an alternative, will be facing significant initial capital investment for equipment upgrades.

The primary question has to do with the opportunity cost of waiting until the standards debates are settled.  Clearly, in the case of the VPLS there is a fertile market, given the popularity of Metro Ethernet deployments.  This could result in late adopters to VPLS to compete for remains of a largely captured market segment. Services Providers that can effectively leverage this technology are those that have a large national Metro Ethernet deployment, such as Time Warner Telecom with 44 Metro Markets throughout the United States.

The future of VPLS looks very promising. The technology offers the perfect extension to the service provider’s existing Metro Ethernet portfolio. The focus of most the vendors should be the Operations and Management (O&M) of this technology. Sending Layer 2 frames across a Layer 3 infrastructure offers a lot of new challenges for diagnosing and troubleshooting customer issues. As far as the technology is concerned, there are several avenues over which VPLS can travel such as Inter-Autonomous System VPLS, TDM (EoTDM) into VPLS and even encrypted layer 2 tunneling over IP from another provider into the VPLS Cloud. One of the strengths of Ethernet throughout the years has been its overwhelming ease of use. Everyone knows how to plug in a hub. Now we are extending the success of Ethernet to sites thousands of miles away. Watch for VPLS to make it happen.

Henry Yu is a Lead IP Architect for Time Warner Telecom.

Purchase reprints of this article by calling (800) 290-5460 or buy them directly online at www.reprintbuyer.com.

Respond to this article in our forums!