Data center interconnect using VPLS

Virtual private LAN services offer an alternative to traditional WAN links and solve enterprise data center interconnect problems. VPLS also opens the door to new data center network management strategies.

Interconnecting multiple data centers is a necessity for large organizations. Companies build multiple data centers to ensure backup, disaster recovery, or failover support for the primary site, as well as to load-balance traffic or simply bring the applications and data physically closer to end users. But supporting these business needs across the WAN can be a challenge because of bandwidth limitations and complex routing. Virtual private LAN services offer an alternative, however, by flattening the network and making the distance between sites transparent to the data center. VPLS may not only be a cure for managing multiple data center networks but could also open the door to taking full advantage of a virtualized environment.

What is virtual private LAN service?

Virtual private LAN service (VPLS) is a telecom carrier-provided service that offers layer 2 Ethernet connectivity between enterprise locations. Unlike traditional WAN connectivity or IP-based virtual private networks, VPLS uses multiprotocol label switching (MPLS) to create the appearance of a virtual private network (VPN) at each data center. A VPLS can provide point-to-point and multipoint services, as well as any-to-any capability. In effect, VPLS creates a virtualized Ethernet switch at the service provider, linking multiple remote sites as if they were on the same physical switch. While initially provisioned by the service provider, configuration and management of the VPLS can fall either to the service provider or to the in-house networking team, based on the needs of the individual customer.

VPLS and its role in data center interconnect

There are a number of inherent benefits to using VPLS to connect disparate data centers. One of those benefits is that VPLS functions on a flat network model, which simplifies deployment and management of the network. Unlike a typical layer 2 IP-based model that most WAN services use, a flat network model operates on the layer 2 data-link of the OSI model. This makes VPLS present itself like an Ethernet switch. It also allows network administrators to build the higher-level networks they see fit, rather than planning their networks around specific IP addresses or hub-and-spoke WAN routing.

VPLS works on a mesh network design. This means that it works as a virtual LAN switch, providing a direct any-to-any connection between sites, rather than the hub-and-spoke model used by older WAN solutions. A mesh design offers two key benefits that are appealing for data center interconnect: single-hop links and resiliency. By offering a single-hop, direct access between remote sites on the VPLS, latency issues common in hub-and-spoke designs are eliminated. Without the overhead of running inter-site data traffic with a router at headquarters serving as a go-between, VPLS offers performance improvement for tasks such as data center synchronization and backup processes. Similarly, without the single point of failure of a centralized WAN hub, a VPLS network is more resilient to potential outages. In the event that one data center loses its network link, the rest of the remote sites can go on unaffected.

Like most physical Ethernet switches, VPLS solutions also offer class of service (CoS) options, enabling the network administrator to configure VLAN tagging and define priorities for each.

Effects of virtual private LAN service on virtualization

The layer 2 nature of VPLS provides additional benefits in virtualized data centers. In order to support the movement of virtual machines between physical machines, technologies such as VMware's vMotion require both the source and destination services to be on the same layer 2 domain, allowing the virtual machine to move between the two physical servers without requiring a change to its IP address. With multiple data centers connecting on a flat, layer 2 network, VPLS enables vMotion moves not just between the physical servers under the same roof but also moves between geographically dispersed data centers. Taken to a logical extreme, virtual machines and their applications can follow the sun, moving to the data center physically nearest to the end users and support staff that utilize them during a particular part of the day.

Getting started with VPLS

Like any other WAN service, working with your telecom vendor for pricing and provisioning will be required for VPLS service. Once deployed, the service presents itself as a simple Ethernet port at each remote site. Most providers offer a Web-based dashboard to allow administrators to monitor port utilizations, adjust VLANs and CoS configurations. Specific functionality and control of the VPLS will vary, however, with both the service provider and the unique customer deployment. Likewise, responsibility for the management and monitoring of the VPLS could vary between organizations, with some enterprises keeping management of the virtual private LAN services within the WAN support group while others hand responsibility for it to the data center network team.

This was last published in September 2010

Dig Deeper on Data Center Networking