A wide area network connects the far-flung employees of North Slope Borough, Alaska, but until recently traffic over the WAN’s satellite links moved too slowly to make sending e-mail practical, never mind sharing large files.
That all changed when the borough encompassing eight villages along the northern tip of Alaska installed a WAN optimization device at each location. Files that previously took about 40 minutes to transfer now take five to eight minutes, says Sean Harty, network administrator.
“Before it was painful, and now people can actually use it,” Harty says about the optimized network. “I could probably cut my bandwidth in half now and nobody would notice the difference.”
Harty deployed Citrix Systems’ WANScaler, one of the many WAN optimization tools gaining popularity as cities and states seek ways to boost network performance as they centralize applications.
Rather than investing in more bandwidth, many agencies choose to deploy application acceleration technologies that eliminate the latency and protocol “chattiness” that occur when standards designed for local networks are used over the WAN. Solutions usually require hardware on both the sending and receiving ends.
Typical WAN optimization tools tap a variety of techniques to handle traffic, including compression, caching, protocol acceleration and bandwidth management. Compression condenses data patterns into reference points on one end of the network and restores them on the other end, making them particularly useful when large files are repeatedly sent over the WAN with small changes. Caching stores repetitive data sequences near the user, proving handy when identical files are sent frequently. Protocol optimization modifies standards that were not designed to go the distance on the WAN to more WAN-friendly standards (see sidebar). Bandwidth management allows high-priority or time-sensitive traffic, such as Voice over Internet Protocol, to move ahead of lower-priority traffic.
Experts caution that, before investing in WAN optimization gear, you must understand the network and identify the cause of traffic delays. Tools for analyzing application performance not only help determine which optimization technologies to deploy, but troubleshoot problems on a regular basis, says Bojan Simic, an analyst with the Aberdeen Group.
“You need to know where your bottlenecks are, and you need to know where to expect bottlenecks when you roll out new applications,” Simic says. “A lot of organizations are jumping into accelerating their WAN traffic and optimizing the applications, but they don’t have basic visibility into the network.”
With the Citrix technology, the most advantageous combination of optimization techniques are automatically applied to each data stream. North Slope Borough spent approximately $60,000 on its WANScaler deployment, which included a backup unit, Harty says. He estimates the technology will pay for itself in six months. Because the WANScalers optimize the borough’s traffic and compress data, Harty is able to reduce the bandwidth of his satellite connections to save money.
Escalating network costs and the need to accelerate business processes and improve customer service drive adoption of WAN optimization, according to the Aberdeen Group.
While often-insurmountable stretches of snow and ice isolate North Slope’s eight offices from each other, vast expanses of desert separate users at the Nevada Department of Corrections. Dan O’Barr, systems and communications manager, needed a way to speed the delivery of applications to remote sites over latency-prone satellite links. Increasing bandwidth was not an option.
Before O’Barr deployed WAN optimization gear from Blue Coat Systems, the department’s inmate tracking system screens took as long as 50 seconds to update — much too long for corrections staff. With Blue Coat’s ProxySG devices, which combine compression, protocol optimization, traffic prioritization and two forms of caching, the screens take about five seconds, he says. Object caching stores files, web content and videos on a local server, while byte caching locally stores repetitive strings of data found in the byte stream.
On average, applications operate three to four times faster than they did before, and some work as much as 15 times faster, O’Barr explains. The department spent approximately $250,000 on 21 devices. The options of doubling T1 lines would have cost about $6,400 per month, he says. “We all feel it’s more than paid for itself.”
The Virginia Department of Motor Vehicles did not have to rely on latency-prone satellite links to connect its 73 sites, but it did want to eliminate bottlenecks in e-mail, file-sharing and desktop applications. After all, sitting around and waiting for applications to open isn’t an efficient use of employee time. To solve the problem, Nick Danforth, information systems manager at the department, installed WANJets from F5 Networks.
The WANJets, which provide advanced compression, data reduction and transport protocol optimization, cost a couple hundred thousand dollars to deploy, Danforth says, but adding another T-1 line would have cost approximately $500 per month per site in addition to the expense of upgrading the infrastructure.
“Our bandwidth costs are recurring, and any cost to increase it would have been exponential,” he says. “We would have had to buy new routers and the new hardware to go along with it. With the WANJets, we were able to capitalize on our existing infrastructure.”
Designed for smaller networks, trusty protocols such as Common Internet File System (CIFS) and Transmission Control Protocol (TCP) don’t always play nicely on the WAN. That’s where protocol optimization can help. Myriad special protocols are used to accelerate traffic between optimization devices. Some exist in the public domains while others are proprietary.
WAN-friendly protocols for optimizing TCP include Fast TCP, HighSpeed TCP (HSTCP) and the Space Communications Protocol Standards for TCP (SCPS-TP). Engineers at the California Institute of Technology’s Netlab developed Fast TCP to offer an alternate congestion control scheme. HSTCP modifies the congestion control mechanism for use with large congestion windows. And NASA and the Department of Defense developed SCPS-TP to allow communication over challenging environments.