In a boisterous over-capacity and overtime session at Interop here today, vendors debated the merits of WAN and application acceleration and whether transparency is required for them to actually work within enterprise environments.
Moderator Erick Siegel, a senior analyst at Burton Group, began the session with a long explanation of why WAN acceleration is necessary. More and more enterprises are centralizing applications and services and are delivering services over the WAN.
"The problem is that the applications work well on the LAN," Siegel said. "But when moved to a WAN, response time is longer than expected and app timeouts appear."
Siegel said there are four key reasons why most WANs are in need of some form of optimization or acceleration.
The first is transmission latency. Simply put: You cannot change the laws of physics.
On a LAN, which is by definition local, an app may have 0 milliseconds latency. When you go from New York to San Francisco latency, goes to 30 ms; from New York to London it hits 50 ms.
"The longer the string the longer it takes, " Siegel said.
Bandwidth restrictions of WAN pipes is also a cause for lag.
"Effective bandwidth depends on the error rate and the ability to fill the transmission channel," Siegel noted. "TCP/IP works in bursts of data and then silence."
Ping-pong latency is also an issue, according to Siegel. The "ping pong" effect is caused by apps that have a lot of back-and-forth activity to perform their functions.
There are a number of ways that vendors are working on accelerating WAN traffic. Among them is file caching, which helps improve response time and decrease bandwidth.
Siegel also identified compression as being a solution, though he noted that you cannot compress encrypted data.
Overall data reduction is also a key technique being employed by app acceleration vendors.
"What they do is build a dictionary at both ends of the path and then substitute pointers for data strings," Siegel explained.
Vendors are also working at protocol acceleration, which improves flow and error control and helps to reduce ping-pong behavior.
Despite the apparent benefits of WAN and application acceleration, there are a number of concerns that enterprises have with the technology.
David Hughes, CTO of Silver Peak, noted that transparency is an issue.
"Whatever the IP addresses are on the client side, they need to appear again on the server so intrusion protection systems (IPS) don't get confused by addresses that change."
A lack of transparency can also be a cause of WAN routing issues.
"It's very important that packets get to the appliance on the other side so they can get decompressed," Hughes noted.
Liad Ofek, vice president of technical services at Expand Networks, also noted transparency as potential issue for enterprise WANs
"In MPLS (define), you need to have certain visibility/transparency to take advantage of advanced services," Ofek said.
Mark Day, chief scientist at Riverbed Technology, went so far as to state that the only way to maintain complete transparency is by not optimizing.
The remainder of the panel members didn't agree.
Dan DeCasper, CTO of Orbital Data, which is now part of Citrix, explained that his company tries very hard to maintain transparency and, for example, doesn't change TCP/IP headers.
According to DeCasper, Citrix has won many deals based on that capability.