In all my musings on WAN optimization and application acceleration, one common factor remains: the networks in need of optimization tools and enhanced traffic speeds rely on TCP/IP as the transport protocol.
This week, in response to a pair of newsletters published on Gartner's take on the application acceleration market going forward, a new vendor (or rather one that I had yet to discover) emerged from the virtual woodwork to share with me its approach to traffic acceleration and network optimization. The informal e-mail exchange introduced me to the company and raised the discussion: What if networks relied on a new protocol? Would today's top concerns around application traffic acceleration fall by the wayside?
Data Expedition, Inc., or DEI, according to its Web site is a privately held company working on network speed and reliability issues. The company, according to one of the business development managers, has developed a "high-performance protocol called MTP/IP (Multipurpose Transaction Protocol for the Internet) that uses data models, flow control algorithms and sophisticated congestion control and error recovery to achieve" performance gains for customers. Gartner concluded in its paper that the current raft of offerings don't attempt to "modify the TCP stack" as a means to improve performance. And DEI says it has based its business on helping customers "overcome bandwidth inefficiencies created by TCP" with its MTP protocol.
"Our MTP/IP technology takes a much more conservative and informative approach and observes network behavior to determine the maximum sustainable data rate of the current path and constantly adjusts as network conditions change," wrote Mike Jackson, a business development manager at DEI, in an e-mail. "No oscillation of information, like TCP/IP, but a more conservative algorithm approach produces a smoother data flow, allowing for previous wasted bandwidth to be recovered."
As the question was put to me, I will now pass it along to Network Optimization readers. Do the developers and engineers at DEI have something here? Or as Jackson asked me, "Would embedding a high performance protocol that bypasses the inefficiencies of TCP, but has the same functionality but much more reliable and faster, into the router operating system be the way to solve everyone's problems?"