The announcement by Expand Networks that it's adding WAFS and CIFS capabilities to its WAN accelerators highlights the growing convergence and consolidation in this area. Increasingly, single-function devices are being replaced by multi-function appliances designed to be all things to all men - or rather, all things to all branch offices.

The aim of most of them is server consolidation: by enabling users in remote or branch offices to access files or applications over the WAN, but at LAN speeds, they allow organisations to bring more IT back to the centre. That means servers can be managed and backed up more easily, or services consolidated onto fewer machines, reducing the costs and support loads.

Yet problems remain and would-be users need to be well aware of them, whether they're technical issues such as the question of what happens to remotely-accessed applications if the WAN goes down, or the organisational challenges highlighted by Gartner analyst Neil Rickard.

The new aim is "one platform, multiple services," says Adam Davison, Expand's EMEA VP. He adds that previously the company focused on WAN compression, caching and monitoring. Then it added plug-ins to optimise for specific applications, including Citrix, and now it has added WAFS too.

"We have a WAN-out mode, and can deliver cached files in read-only mode and then synch when the WAN comes back," he notes. "But if you remove the fileservers from branch offices, you also lose all the other fileserver services - printing, DNS, DHCP, etc. So we deliver those on our device too - that's very important for large users with thousands of sites."

Meanwhile, rival supplier Riverbed Technology highlights academic research which has formally proved something that most of us only believed intuitively: you can't have everything. If the WAN goes down, then you must give up either data consistency or data availability.

That means that if you use local caching, then either you have to make the cache read-only and lose full availability, or you run the risk that two users will try to update the same file and someone will have to sort out the inconsistency - probably manually.

There can be other consistency problems too. For example, if a file is opened via a remote device, some systems will lock it on the central server to maintain consistency. However, if the WAN or the client device should fail while the file is locked, or if the server fails or reboots, writes could be lost even though the user thinks they have been made successfully.

"The nature of caching is you're trying to do operations locally then fix things up in the background," says Mark Day, Riverbed's chief scientist. "The drawback is where devices get damaged or you get disconnections, you have the potential for file corruption. If you buffer writes locally, you don't have an answer for what happens if a box can't resync, or if multiple boxes write locally.

"The vendors have to declare that they're done at the point where they declare there's multiple copies and leave the user to get on with it. They say there's no data lost, but the user may not have the tools to fix the problem.

"We make it explicit to the customer - there's no free lunch. Where do you want control of your files - in the branch or at the centre? If your situation is that disconnection is not an issue, global mode is great."

Mixing it all together
Everyone has this same problem though, and the only difference is in how they present it - or in some cases, in how they try to hide it. More to the point, WAFS and its equivalents have become just one ingredient going into the finished pudding, the proof of which will be in its eating.

"All the suppliers are converging, then for the customer it's ease of use, ease of maintenance and so on," says Expand's Davison. "The key is transparency - the ability to put it in without touching the other devices in the network.

"There's still a lot of education going on," he adds. "Every site is configured differently, so each is a unique challenge." He suggests that network and application acceleration may ultimately belong with telcos or managed service providers, as an appliance alongside an existing managed network.

"It's a complicated landscape," agrees Alan Saldich, Riverbed's product marketing VP. "A year ago there were three categories of products - WAN optimisers such as Expand and Peribit, file caches, such as Actona, DiskSites and Tacit, and application acceleration - Netscaler and co.

"There's been a lot of acquisitions in the last year, and we're now seeing the development of a superset of products or features, aimed at solving the whole remote office problem. The new term is wide area data services - it's all converging. People want to fix the entire problem, which is that in the real world, bandwidth is not free."

Saldich stresses though that there are still technical differences to be aware of. For example, he says that WAFS is "just file caching, and file caching doesn't do much for other traffic. The data we store on disk is application-independent, in highly granular form - it's not files. It gives the ability to use it for any application running over the network, not just file access."

He warns too that "there's a big difference between web applications and other applications, but some web application acceleration companies leave the 'web' off. A chatty application probably requires application-specific knowledge - we address that because we're a TCP proxy, and TCP issues are applicable across varieties of traffic."

Adam Davison agrees that application-specific knowledge is vital: "We started with application plug-ins about a year ago, using features such as TCP fast-start and adjusting the window size," he says. "It's not just about compression any more, it's all about giving the best user experience."