Meeting the WAN Challenge

The ultimate goal of all network infrastructure is to deliver applications effectively to users. Delivering applications effectively in the local-area network (LAN) is easy; preserving that performance across the wide-area network (WAN) remains challenging.

To ensure enterprise-class application delivery across the WAN, IT must address four key areas:

Bandwidth The most obvious restriction for application delivery over the WAN is reduced bandwidth. Businesses today run more applications across the WAN, and often the bandwidth requirement per application has increased. Web-enabling applications such as ERP systems can cause a transaction's bandwidth to increase as much as tenfold compared with client/server architectures.

An alternative to upgrading WAN links is to apply data compression to reduce the traffic consumed by applications. For example, repeated data sequences in a file, such as repeated letters, words, or images, are replaced by a short flag instead, so a file that would have consumed 1 Mbit/s can instead require only 250 kbit/s or even less.Latency – The impact of latency has historically been a little less obvious and less well understood than bandwidth limitations. Sometimes latency doesn't just degrade application performance; sometimes it limits the overall application throughput or breaks the application altogether. Enterprises that have purchased sizable WAN links often assume they are protected from performance problems, but latency can limit throughput regardless of bandwidth.

Reducing latency itself is not possible, but organizations must limit its impact, especially on TCP-based applications. This is important because TCP (Transmission Control Protocol) relies on a series of back-and-forth transmissions to make sure that the data is properly received. Applications waiting for that back-and-forth traffic cannot make full use of the WAN bandwidth and are slowed down by latency.

To address latency, organizations need techniques that allow older systems to send more data per transmission, eliminate the round trip time (RTT) from the TCP session startup, and replace TCP with another reliable, but more efficient, transport.

Quality of Service – Once applications hit the restricted WAN bandwidth, they must contend with each other for access. To make sure the important applications get access to the link, organizations need simple graphical user interfaces (GUIs) to set up bandwidth allocation and QOS. Historically, QOS tools have required complicated, text-based configurations to be enabled as well as detailed information, such as pattern flow data, that IT staff may not have. A wizard-based GUI simplifies the process by asking IT staff simple questions such as which applications matter in which locations and then setting the parameters on the equipment as needed.

Path Optimization – Public Internet services provide compelling cost savings, but poor Internet performance has limited their use. Companies need a way to gradually migrate less-critical applications to public links while keeping a close eye on link performance.To achieve this, organizations need to focus on certain key areas, including designating which applications should flow over which link and defining acceptable latency and loss metrics. It is also important to define the actions the WAN infrastructure should take if performance drops below those levels.

These tools for addressing application delivery – increasing bandwidth, reducing the impact of latency, enabling application prioritization, and selecting the best path for applications – are critical if IT is to provide acceptable performance levels for employees. In today's demanding business climate, anything less will leave companies waiting – and lose customers in the process.

— Michelle Rae McLean, Director, Peribit Networks