Cisco IT’s global WAN has a two-tiered architecture. Tier 1 is a DWDM/SONET/SDH-based global backbone that connects 22 Cisco facilities and WAN aggregation hubs in 11 countries. Tier 2 regional WANs connect at the hub sites. The regional WANs are built based on the most appropriate technology for the region, including MPLS VPN in several locations. In North America the regional WAN is currently a TDM-based private-line network with eight WAN aggregation hubs in the United States and one in Canada which we recently evaluated for possible migration to MPLS VPN.
Our evaluation highlighted the fact that MPLS VPN service offerings and cost models vary from region to region across the globe. Conditions that make MPLS VPN the WAN technology of choice in one part of the world do not necessarily apply elsewhere. In 2001, when we reviewed our SONET/ATM/Frame Relay/TDM-based WAN in Europe, we found that a service-provider-based MPLS VPN network solution was the best option for us. We were able to increase our WAN bandwidth by more than a factor of four for about the same cost. We were also able to meet our requirements for QoS for real-time voice and video applications, and native multicast needed for streaming Cisco TV broadcast of companywide and other meetings. (We support seven QoS classes, as described in Cisco IT case studies on QoS and Medianet.)
Overall, we’re extremely happy with our MPLS VPN network in Europe. Since it was built we’ve expanded our use of MPLS VPN into Eastern Europe (Russia and CIS countries) and again into the Middle East.
Because of our earlier experience in Europe, we considered deploying MPLS VPN to try to realize similar cost-savings benefits in North America. In 2001, when we first looked at MPLS VPN for North America, we struggled with supported features (such as native multicast) and service provider coverage. In recent years those issues have largely disappeared. More recently we have gone through a costing exercise to determine how we can meet the needs of the business in the most cost-effective manner. In evaluating costs we not only looked at the cost for the MPLS VPN service itself but also the significant cost of the resources to migrate the network to the new technology.
Over the years we have evaluated MPLS VPN for North America on several occasions and have issued three separate RFPs, but we have never found the substantial cost-savings that we found in Europe. Two main factors explain the difference. First (and by far the more significant), is the fact that in North America we deal exclusively with domestic circuits. The domestic circuits are relatively inexpensive because of our backbone and nine hub locations which keep the average circuit distance to about 200 miles (TDM circuits are priced by mileage). In Europe you have a large concentration of different countries within a relatively small geographic area. Connecting sites in many countries back to WAN aggregation hubs in different countries via International Private Line (IPL) circuits is quite costly. MPLS VPN with in-country Provider Edge (PE) routers eliminated the need for the expensive IPLs, which were replaced with much less costly domestic circuits. Second, Cisco’s rapidly expanding usage of voice and video applications such as WebEx and TelePresence require a large amount (approximately 70 percent of the average WAN DS3 link) of expensive guaranteed bandwidth from the service provider. With MPLS VPN in Europe the cost of that guaranteed bandwidth for voice and video is more than offset by the elimination of the expensive IPLs. That opportunity did not exist in North America.
With that said, I do not assert that MPLS VPN in North America won’t work for others. Whether the technology is appropriate or not depends heavily on your specific requirements at the time of the evaluation. The marketplace (including the cost models) for MPLS VPN continues to evolve and Cisco IT will consider MPLS VPNs as a possible future alternative as we periodically reevaluate all our regional WAN strategies.