Few can have failed to notice the huge range of network and application acceleration products now available. From Zeus' ZXTM, which won best accelerator product in the 2005 Techworld awards, to Peribit and Fineground, which were snapped up earlier this year by Juniper and Cisco respectively, acceleration is hot news.
And the topic is not about to go away, according to Neil Rickard, a vice president at market research company Gartner. Instead, he believes acceleration technology is going to dramatically change the way we build and use networks, websites and applications.
"The application-fluent network, where the network speaks the language of the applications, is forming a major focus for us now," he says. "The growing portfolio of application optimisation is becoming a new product class."
He acknowledges the tactical potential - saving bandwidth to defer a leased line upgrade for example, or improving application response over the WAN. "The TCO for WAFS is very strong - the support and software costs for remote servers are typically very high," he says. "You get improved security and backup from it too, and there's all sorts of soft issues such as improved regulatory compliance."
But he stresses that network and application acceleration is much, much more than that. "It can change how you deploy applications," he says. "For example, WAFS lets you take the file and print servers out of a remote office. Or with web and rich-media caching, you can change how you use e-learning. With one of these boxes in place, you can re-visit almost all your applications and look at how to deploy them better.
From tactical to strategic
"Historically they were point products to fix specific problems, such as poor Web application performance. Now we're increasingly urging clients to think of these as a layer in the network, as a strategic platform. Sure, you have to weigh the tactical issues too, but think strategically."
Of course, there are many different manufacturers to choose from, but the technologies are starting to converge now. Two classes of device are emerging - WAN optimisation controllers that work in pairs, and single-ended server load balancers or application delivery controllers from the likes of F5, doing SSL offload, DDoS mitigation, content reformatting and so on.
Rickard says that the functionality leaders at present are probably Expand and Peribit: "They overlap substantially, but they have slightly different product sets. Then there's Packeteer and the others.
"The problem for Cisco is it still has major holes, for example it's weak on WAN optimisation - the areas that Expand and Peribit are doing. Cisco has subsystems that do most of it, but to actually build a working system you would have a mountain of blades and boxes, many of them with their own management platforms too. Given a couple of years though, Cisco can integrate all those systems."
The technical selection will depend on how long you think the manufacturers will be around, and on what you want to run over your application-fluent network. "There can be enormous differences between organisations in their applications and how they're deployed," Rickard says.
"We'll probably all have email, financial accounting and so on, but not all will use FTP or videoconferencing, say. It's tricky - you do need to match accelerator systems to your environment and see how they'll work with your applications."
Blending the network layers
However, the real challenge facing network designers is independent of the manufacturer. It's that the layers of the network are no longer insulated from each other - if they ever were - yet organisations have evolved to mirror the OSI seven-layer model, with one department managing the apps layer, another responsible for the IP network, a third for the physical layer, and so on.
Acceleration devices break those borders down completely, so those groups need to communicate on all sorts of levels. For example, the business plan for an infrastructure investment could mean savings for the server and applications groups, but if the organisational structure means there's no benefit in that for the infrastructure group, it won't happen.
Conversely, it can affect how you build applications too. You could build a website in a highly cacheable form, or in a way that's not cacheable at all, but do your developers think about the presence of caching in the infrastructure?
"You need to fundamentally change the behaviour in the network department," Rickard says. "People changes are harder to make than technology changes - plus, you have to realise those people changes are needed in the first place."