Virtualizing WAN Optimization

Sprint

The number of applications that can’t run in a virtualized environment is rapidly dropping. Earlier this month Mitel announced its new focus on enabling call control and UC applications to run on general purpose VMware hypervisors. Now, WAN optimization controller (WOC) vendors are also delivering virtualized support for their applications, further reducing the need for stand-alone appliances.

Running WAN optimization on virtualized servers enables IT to use the same servers that it uses for other essential branch services (DNS, DHCP, print services, etc.). This eliminates the need for a separate box in each location, if there is already adequate hardware in a typical branch, and thereby reduces administrative costs and reduces the weight of the deployment. Virtual appliances are easier and faster to deploy, and to move, and they offer greater options for resilience and disaster recovery.

If there is not adequate hardware on site already to handle the optimization load (e.g., no server in the branch, or only servers with insufficient processing power or memory, or inadequate network interfaces), then IT will have to deploy new or beefier hardware. This means increased capital outlay to make sure the virtual WOC can do its job. (And of course, if there is no server there yet, then IT is just adding another box to the site, just as if it were adding an appliance; although on the plus side, it is not a “unitasker” and IT can use it to deliver other services, too.)

The virtual appliance approach also allows some shopping around for best-cost platforms, or, conversely, the ability to use the organization’s preferred server hardware vendor. This last scenario is an option IT administrators should welcome—Nemertes benchmarking finds 50% would be just as happy running a virtual optimizer as a physical one, if performance was equal.

Of course, running virtual appliances even on existing hardware doesn’t eliminate capital costs, as pricing for virtual appliances is carefully gauged to forestall cannibalizing physical appliance revenues. And although it can reduce administrative costs, there remains the need to manage a box (although virtual) in every location, meaning additional management expense.

As an alternative to deploying optimization as an encapsulated workload in the form of a virtual appliance, IT can run optimization as a workload alongside the services being optimized. For example, a Web traffic optimizer can run alongside the Web server on the same virtual or physical server, as a system service rather than a virtual appliance. This approach is often enormously cheaper than deploying appliances (physical or virtual) to branches, but it only optimizes one kind of traffic, such as Web traffic, and can only control traffic to the one server. Proliferating point solutions for different services would lead to complexity and management issues, as well as the likelihood of exceeding the cost of a network-based approach.

WAN optimization appliances constitute a growing market now, but the writing is on the wall: A few years from now "WAN Optimization Controller" may barely exist as a separate category of physical network appliance. As WAN refresh cycles play out over the next seven years or so, Nemertes expects a majority of companies to migrate to one of the other optimization models now available: carrier/cloud services (optimization as a service), point solutions and virtualized optimization appliances (optimization as a workload), or integrated router-switch/optimizer or security/optimizer appliances (optimization as a feature).

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2011 IDG Communications, Inc.