The Return of Layer 2: How Data Center Networking Requirements Will Change the WAN

Feature Article

The Return of Layer 2: How Data Center Networking Requirements Will Change the WAN

By TMCnet Special Guest
Alex Foster, Product Manager, Data and Managed Services at Cavalier Telephone
  |  February 01, 2011

This article originally appeared in the Feb. 2011 issue of INTERNET TELEPHONY Magazine.

For the past five years MPLS-based IP VPNs have been the dominant WAN transport solution for organizations of all sizes. Originally adopted and deployed by large organizations to support interconnection between thousands of sites, IP VPNs quickly trickled down market to the point of ubiquity. The widespread adoption of IP VPN has taken place because it offered a complete solution to a common corporate need – secure, fully-meshed IP connectivity with quality of service between locations.

While WAN architectures have converged on a Layer 3 approach, a quiet revolution has been under way in data center networks and is increasingly creating requirements that IP VPNs cannot practically meet. Modern data center networks are designed around the requirements of virtualization and SAN replication. One of the biggest requirements for fully leveraging virtualization is that virtual machines be able to move between physical hosts while maintaining their IP addressing. Virtual machine movement within a data center is already common and has driven a reinvention of data center switches and LAN architecture. Virtual machine mobility between data centers for disaster recovery is the next frontier, and the networking requirements around this technology and SAN replication will drive a new set of connectivity requirements not easily met by IP VPNs.

Thus far vendors and network architects have responded with several strategies for accommodating these needs within the existing WAN architecture. The well-tested solution is simply to operate outside of the IP VPN and deploy SONET or WDM connectivity between the data centers to act as a LAN extension. This solution is costly, does not scale with several data centers, and it may be difficult to even get connectivity between data centers if they are served by different carriers.

A more recent solution to extending Ethernet between data centers is to overlay Ethernet traffic on top of IP VPN networks using tunneling to encapsulate Ethernet frames into IP packets and then strip them back out at the far end. This solution eliminates the operational expense of running parallel networks but requires a substantial hardware investment and a significant amount of expertise to configure and maintain.

Both solutions are far from ideal. The preferred solution for most mid-size businesses increasingly will be carrier-delivered Layer 2 VPNs. This category includes services such as virtual private LAN service, E-LAN and transparent LAN services. These services are delivered over the service provider’s MPLS backbone but leverage functionality in the provider’s edge to deliver a Layer 2 service based on virtual bridging rather than a Layer 3 service based on virtual route forwarding. This gives the customer a network that behaves like a wide area Ethernet switch. For many companies this single Ethernet WAN can serve as a solution to interconnect both data centers and offices.

VPLS allows customers to solve many of the common data center interconnection problems in a much simpler and more intuitive way than in an IP VPN environment, and is far more affordable than SONET/WDM solutions. Without having to peer with service provider routers at Layer 3, subnets and Layer 2 domains can easily be extended between data centers. With complete control of the IP layer, routing can be based on the preferred protocol rather than a limited set of protocols supported by the service provider. Perhaps most importantly, organizations can tune their routing protocols to deliver convergence times far quicker than possible in an IP VPN environment. Finally, many technologies that have a long history of deployment in data center environments such as hot standby router protocol can be leveraged in the WAN for failover between data centers with complete IP mobility.

Although there is concern in some corners about latency over the MPLS core, these concerns are generally unwarranted. The latency imposed by MPLS hops in the carrier core is on the order of 12 to 50 microseconds, nearing irrelevance even in a world of single-digit millisecond requirements.

While VPLS and E-LAN solutions are not yet common, they are growing rapidly. Virtually every major service provider has announced some form of VPLS offering, and customers are quickly realizing that removing the service provider from the IP layer of their network can greatly simplify their lives and speed the deployment of next-generation data center solutions. The instances where separate corporate and data center networks are justified will continue to decline, and the likely winner in these cases will be Layer 2 VPN solutions such as VPLS.  

 Alex Foster is product manager, data and managed services at Cavalier Telephone

TMCnet publishes expert commentary on various telecommunications, IT, call center, CRM and other technology-related topics. Are you an expert in one of these fields, and interested in having your perspective published on a site that gets several million unique visitors each month? Get in touch.

Edited by Stefania Viscusi


Sign up for our free weekly Internet Telephony Newsletter!

Get the latest expert news, reviews & resources. Tailored specifically for VoIP and IP Communications.