WAN optimization in a large enterprise is a nascent and headline-grabbing market. Still, IT managers are faced with the burning question: Is it really worth the investment?

The technology uses techniques such as compression, caching and de-duplication to optimize traffic, without actually making the networks run faster. Whether it's a panacea for network ailments is open for debate, and analysts are at turns optimistic and mildly skeptical.

"WAN optimization addresses two classes of problems -- scarce bandwidth and high latency," says Joe Skorupa, a Gartner Inc. research analyst. "This often reduces the average number of bits transmitted by more than 90 percent [for data patterns that have been seen before]. However, this can be misleading, since new data patterns are reduced by around 50 percent, and compressed or encrypted files see no reduction."

Skorupa continued: "[Providers] add QoS features to deal with bursts so high-priority traffic can get through. Yet, even in these cases, link loads that consistently top 65 percent to 75 percent cause problems that typically require bandwidth upgrades." He said this is because networks running at an average of 65 percent to 75 percent capacity often have short bursts that go well beyond that level. That leads to unpredictable response times because there is no spare capacity to absorb the bursts.

WAN optimization to the rescue

Specialized appliances from companies such as Riverbed Technology and Citrix Systems address issues such as unresponsive apps, slow transmissions and network congestion.

Managed application providers such as AT&T and BT Global Services perform the same duties and charge a fee for carrier service for the network pipeline that connects remote offices and retail stores, for example.

Interestingly, there are dramatically different reasons why large companies would choose to optimize WAN traffic, depending partly on their markets, infrastructure and security concerns.

For example, at Solutia Inc., a chemical manufacturer in St. Louis with about 6,000 employees and annual revenue of about $4 billion, an application performance problem was a result of quick expansion, ongoing server and data center consolidation efforts, and what looked remarkably like start-up costs -- big upfront payments for new branch offices to access home office services -- to gain more network capacity, a peculiar concept for a 100-year-old company.

For more enterprise computing news, visit Computerworld. Story copyright Computerworld, Inc.