There are actually many options in actual practice today when it comes to the need for connecting geographically dispersed Data Centers. Obviously, the geographic dispersion is highly favorable for data availability, and the interconnection technology that is chosen should meet the business needs of the organization as well as the generic technology requirements.
This article examines a Cisco solution available in the Nexus line known as Overlay Transport Virtualization or (OTV). The purpose of this document is to provide an overview of the technology. Much more detail is provided in the CCIE Data Center Written Bootcamp here at IPexpert.
Many Data Center designers applaud Cisco’s OTV as it is a very clean design. In this solution, Data Center traffic is bridged directly over an IP core with no required intervening technology such as Multiprotocol Label Switching (MPLS) or pseudowire.
This technology does not feature any type of dynamic MAC address learning. Instead, control plane MAC address reachability is advertised using the IS-IS (Intermediate System to Intermediate System) routing protocol. Broadcast and multicast traffic is not replicated and sent through the core transport, but instead, native multicast is used to transport such traffic.
Another exciting point for OTV is the fact that Spanning Tree Protocol (STP) is not used at all by the technology. Instead, STP is limited to individual Data Center sites. Obviously OTV still requires loop prevention, and as such, the technology uses a simple global split horizon function built into the protocol mechanics.
Finally, to round out the robust and desirable feature set, OTV features automatic multihoming behavior for nodes allowing elegant per-VLAN designated forwarding devices. The failover, therefore, is automatic. Note that two OTV capable devices are required in each data center that requires OTV connectivity.
The two devices in the data center that are running OTV are known as Edge Devices. These devices feature Layer 2 ports facing the data center, and Join Interfaces facing the IP transport cloud. These Join Interfaces are Layer 3 routed interfaces that behave like host interfaces. They possess an IP address, but do not engage in dynamic routing protocols.
Another component in the OTV design is the Overlay Interface. This is a simple configuration mechanism requirement and is a virtual interface that carries the actual OTV configuration. It is a logical, multi-access, multicast capable interface that does not participate in Spanning Tree Protocol (STP).
A Site VLAN is configured between the Join Interfaces. This VLAN must be dedicated for OTV and is used for such mechanisms as the forwarder decision process.
The OTV implementation also features an Address Resolution Protocol (ARP) Name Discovery (ND) cache. This ARP snooping device helps to eliminate site-to-site ARP functions.
In summary, OTV is an excellent option for Cisco Data Center connectivity that features an extremely simple configuration of pure MAC-over-IP transport. The technology solves a major issue with competing technologies that is the elimination of unknown unicast flooding behavior in the cloud. This technology is recommended for implementation using a Virtual Device Context in the NX-OS.
In addition to the IPexpert CCIE Data Center Written Bootcamp, readers should consider visiting Ivan Pepelnjak’s ipspace.net for more information on OTV and related data center technologies.
Anthony Sequeira CCIE, CCSI