FRAMINGHAM (04/10/2000) - It's no secret that today's networks are under stress. Bandwidth growth can't keep pace with demand from increasing numbers of users, as well as an explosion of increasingly complex applications. User expectations for application responsiveness also are increasing.
Adaptive acceleration - a new method of expanding existing bandwidth and improving efficiency - offers a remedy. The technology affords ISPs, application service providers and other enterprise customers a comprehensive, auto-learning technology that provides significant improvement in handling data from diverse applications and protocols.
The result is a boost in network performance and an unclogging of bandwidth bottlenecks. In fact, independent testing has shown that adaptive acceleration can boost bandwidth availability by up to 400 percent.
How it works
Adaptive acceleration is based on several algorithms, including selective caching, vertical data analysis (VDA), adaptive packet compression and quality-of-service prioritization.
Selective caching, an advanced algorithm that identifies which data is worth caching and stores it for later retrieval, is not limited to a specific protocol. Data may include full objects such as an entire .gif file, or a partial object such as a color palette of a .gif file.
Selective caching is effective with Internet traffic (if multiple users go to a certain site, many parts of the site will be cached), and with intranets, where most users are accessing the same site, as well as with Citrix Winframe, SQL traffic, Web casting and Web applications.
VDA dynamically identifies different protocol segments in a packet. Generally, each segment relates to a different part in the protocol stack. For example, VDA might separate a packet into the following sections: .gif file, HTTP/1.1 header, XML, HTTP/1.1 header, TCP, IP and Cisco High-Level Data Link Control.
Separating segments in a "semidynamic" manner, VDA is programmed with basic rules and with understanding of some protocols. Using this preprogrammed understanding, it parses packets into their different segments. "Header style" information in each packet - such as sequence numbers and protocol identifiers - is compressed using a dynamic algorithm.
Adaptive packet compression is used in conjunction with VDA and selective caching to reduce data that is not cached, or which was not handled by VDA.
Adaptive acceleration features an integrated mechanism that can give time-sensitive traffic such as voice over IP higher priority than other types of traffic, moving it to the front of the acceleration buffer for immediate transmission.
Currently, voice-over-IP traffic can be given a higher priority than other traffic types. In the future, other traffic types will similarly be able to jump the queue based on a company's needs.
While "free" bandwidth - whether fiber optic, digital subscriber line or wireless - will afford a certain amount of flexibility, utility costs will remain as certain as death and taxes. Clearly, at least until the promises of bandwidth freedom come to pass, companies will need to find creative ways to maximize existing bandwidth and stretch their infrastructure dollars as far as they can.
Supinski is vice president of marketing at Expand Networks Inc., a maker of bandwidth acceleration products in Roseland, N.J. He can be reached at dsupinski@ expand.com.