Engineering firms know all about taxing the WAN, given the volume of huge design files that typically travel between offices. Strand Associates, a 330-person engineering firm based in Madison, Wisc., has 10 offices and has been accelerating WAN traffic since late 2004, first using devices from Peribit (which Juniper acquired in 2005) and today using Riverbed’s Steelhead appliances.

Justin Marthaler, IT manager at Strand Associates, got in touch after I sent out a call asking for input into the future of WAN optimization. Here’s what he had to say about the latest trends and directions shaping WAN optimization efforts at his firm.

Vendors have been making it possible for companies to add features such as DNS, DHCP and print serving to WAN appliances deployed at the branch. Why is this important? What would you like to see bundled next?

The addition of these services makes it easier to connect an office to the WAN without the need for a dedicated local server. I would like to see Active Directory services added to improve authentication times. Some local storage on the device, tied into our Active Directory permissions, might be helpful. Even though the WAN accelerators speed transport over the WAN, some applications, like Bentley MicroStation or Autodesk's AutoCAD, perform better locally when dealing with large files and large linked or referenced files.

Are client-based WAN optimization technologies on your radar? Why or why not?

These technologies are on the radar, but it's something I'm keeping an eye on, not actively looking to deploy. Most of our users work within a physical office so the appliances we have in place do the job quite nicely. However, as the demand for mobility goes up and we have more people at remote sites without an appliance, I could see an optimization/acceleration client becoming more necessary.

Do you want to see content-delivery features built into WAN optimization gear? What’s driving this trend?

No, this isn't something we'd utilize at this point. However, if there were an appliance or software that could optimize my live videoconference traffic, I'd take a serious look at that. But the H.264 videoconferencing protocol is already quite optimized and WAN acceleration devices don't help live video streams so I don't think there's much that can be done in that area.

Is your team responsible for meeting application-specific SLAs? Or do you expect to be in the future? Do you have the tools to make those sorts of guarantees? If not, what’s missing?

We're not responsible for specific SLA's, but our users let us know in a hurry if application performance isn't up to par. We can use QoS to guarantee performance on some of our applications, but that approach doesn't work for all applications. For applications that run over the same port, our QoS strategy doesn't help us much as we're lumping traffic together based on IP and port number.

How have hosted applications (SaaS), composite applications (SOA) and Web 2.0 applications changed optimization strategies? Do these kinds of applications complicate traditional acceleration tactics?

We're still "old school" in this regard so we aren't heavily using Web 2.0 or hosted applications at this point.

How do existing management features hold up? Do you wish for greater manageability of WAN appliances?

I'm pretty happy with the existing management features I have. I'd like a central management tool that doesn't cost a bundle. We've avoided purchasing a central management device because we have only 10 devices and it's more cost effective to manage them individually than spend the money for the management appliance. The centralized management would be nice, but not at the current price.

What developments in WAN optimization do you think are most important? What features are you holding your breath for?

I think caching was the biggest advance. The difference between the first Peribit we deployed (without caching) to the Riverbed and Juniper devices we tested with caching were dramatic. Everything since then has seemed more incremental, an evolutionary change rather than revolutionary change. I'm looking for dramatic improvements in performance with Bentley MicroStation and Autodesk AutoCAD. I'd also love to see a way for devices to accelerate all data as if it were "warm." We are continually educating our users about the difference between the first time they open or copy a file compared to the second and subsequent times. People open a large file over the WAN for the first time, think it's too slow and then never try it again. After we explain how it works, they get it and are more patient with that first open. I'd like to see some sort of technology that would crawl our network and warm or pre-populate data that is likely to be used over the WAN. I know there are methods to pre-populate data, but those require some pre-planning by the people working with those files. I have no idea how you'd come up with an algorithm that would know what to pre-populate, but it would sure be impressive if it worked.

Ann Bednarz is associate news editor at Network World.