Most companies are attempting to improve uptime through service resiliency. The idea is that when a service fails it can easily recover and resume operation. To improve service resiliency, you might deploy redundant systems so that any failure has limited impact.
Building highly available applications is difficult, involving long project execution times and considerable expense. As application complexity increases, confidence in the service often goes down because it's hard to develop, test and validate the resulting system. This means infrastructure systems that provide application redundancy and resilience are attractive choices.
In terms of OTV and LISP, redundancy means you must have two physically separate data centers that are able to run your applications. Resiliency means that applications can recover automatically from a failure in one data center by transferring service the second data center.
OTV Is Resilience; LISP is Redundancy
OTV provides a way to extend Ethernet networks between data centers with some safety mechanisms for operational integrity. Therefore, a system can move anywhere in an Ethernet VLAN, its IP address is unchanged, and, thus, it can maintain connections to the clients.
By extending Layer 2 Ethernet network between data centers, your virtualization management platform can move virtual machines between data centers. Virtualization services such as VMware High Availability can use this capability to rapidly switch between data centers and provide a resilient service.
LISP was developed by Cisco to allow a system to move anywhere in the network without changing its IP address and maintain client connections. LISP requires the network to be aware of every server, physical or virtual, and to modify network routing as the system and its IP address moves within the network.
Comparing OTV and LISP
OTV is attractive because it delivers resilience without modifying the application and can be deployed with limited impact to the network architecture. The negative aspect of the standard is that the network paths from VM to client will not be optimal because the client traffic must "'trombone" across the Data Center Interconnect (DCI). That is, user workstations accessing a VM in DC1 will continue to use the DC1 IP address even after the VM is moved to DC2. All client traffic must loop from Client-DC1-DC2-DC1-Client. The DCI bandwidth must be large enough to handle this traffic flow and, of course, big bandwidth means big costs. OTV by itself is seductively simple but expensive to implement.
LISP solves the "trombone" effect by propagating routing updates into the access network, including the WAN. When a virtual machine migration occurs, LISP can send routing data to the WAN so that clients will attach to DC2 directly.
LISP is far more powerful than OTV and scales to large numbers of servers. However, it requires significant upgrades to many elements in the network, including the data center switches and edge routers.
You can combine OTV and LISP to deliver a better service. Use OTV to enable VM mobility between data centers and use LISP to ensure that the path from the WAN to the data center is optimal. The two protocols together deliver both service resiliency and redundancy from the infrastructure layer and require little support or integration from the application.
Most companies are choosing to implement OTV today to solve their short-term needs for resilience. At the same time, Cisco has developed new technology with LISP that offers to address the OTV limitations through ongoing investment and planning in the network. It's a complex technology, but Cisco has delivered a strategy to solve these very real problems--in fact, it's a long way ahead of most customers on these technologies. But then, OTV and LISP are “engineer-driven” features, not business initiated.