When organizations are looking for a WAN optimization product, they often start with a simple question: Which one's fastest? My answer: Does it even matter? Performance is important, but it's not where an organization should begin when evaluating hardware and software.
Most mature WAN optimization products deliver competitive optimization ratios on the most common office traffic, such as HTTP and file services. (See my previous blog on the factors inhibiting network throughput.) Sure, there are differences, but they will be plus or minus a few percentage points among the established competitors. Bigger differences occur in the border cases--high rates of loss, unusual protocols, and so on. Performance in these cases is a differentiator, but not nearly as much as it once was.
This shouldn't come as such a surprise; it's endemic to any mature IT technology.
So rather than focus on performance, IT should spend more time examining other issues. For instance, ROI as demonstrated by protocol/application support, pricing, scalability, deployment cost and operational cost come to mind. These issues will be influenced by whether you deploy the product as a virtual instance or in hardware. I'll drill into each one a bit further.
Protocols and Applications: Not every WAN optimization product can accelerate every application or protocol. To maximize ROI, you'll want a product that accelerates as many of the protocols running on your network as possible. List the full range of protocols and applications running on your network today: TCP- or IP -based protocols, encapsulated protocol, proprietary applications, encrypted protocols, all of them.
Don't just think in terms of the corporate-sanctioned applications; look at your logs to find those that may have surreptitiously crept into the fold. They, too, could be consuming precious WAN bandwidth. The more critical applications and protocols that can be accelerated by your WAN optimization product, the better your return on investment.
Pricing: Implicit in ROI is the price tag for WAN optimization. The purchase price is one cost, but there are others, including software licensing. If you're going with a virtual instance, that software has to run on something, so include infrastructure costs, as well. Recurring costs typically take the form of a maintenance contract. The maintenance costs over three years are often as much as, if not more than, the purchase price. Services and some software providers allow organization to pay by the month.
Scalability: While optimization performance may vary little, scalability is another matter. Some products can scale to gigabits of WAN traffic, but not every environment needs it. To estimate how scalable you need the product to be, consider the raw throughput of the product, as well as the number of simultaneous sessions or flows it supports. I'll delve more deeply into this subject in a later post.
Deployment Cost: With enterprises ever more distributed, organizations need to think through what it will take to acquire, test, configure and deploy a WAN optimization product. This is particularly true because as your organization embraces virtualization and virtual appliances, deployment costs are expected to drop. There's no question it's less costly in terms of time and money to deploy a virtual appliance to other data centers and office sites, though it may be easier to rack up a hardware appliance in some locations.
Operational Cost: Operational costs can be tricky to measure. Are there hardware updates or software patches that are necessary? If so, can they be done simply? Most WAN optimization products proxy client-server session to some degree, so outdated protocol implementations can lead to subtle compatibility problems, which may disrupt application communications, and be very, very hard to detect and troubleshoot. Such issues will raise your operational costs.
The Big Picture: How does your WAN optimization product fit into your broader application acceleration architecture? WAN optimization may be all you need, but some performance problems may lie elsewhere in your network architecture. Similarly, do you need your product to fit into a grander strategy, such as an initiative to move to an all virtualized or software-defined architecture? Hardware appliances will work fine for small offices, perhaps, but in sites where virtual software is the norm a virtual appliance might be the only feasible approach.
Performance is an important factor and one that needs to be considered, but it's hardly the most important one when choosing among closely competitive optimization technologies. Focus on the above issues before worrying about which product performs best. Speaking of which, anyone interested in knowing how to overcome loss, latency and bandwidth problems? Next time--I promise.
David Greenfield is a long-time technology analyst. He currently works in product marketing for Silver Peak.
