Massachusetts Institute of Technology (MIT) researchers have developed a computer system called Remy that automatically generates transmission control protocol (TCP) algorithms with transmission rates two to three times higher than those designed by humans.
TCP regulates the rate of computer transmissions to minimize network congestion, and human engineers have worked to improve TCP congestion-control algorithms over the past 25 years. Remy is a machine-learning system that tests variations on possibilities that seem to work best. Certain network characteristics, such as whether the number of users changes and by how much, are determined by users. In addition, users establish network performance evaluation metrics and a traffic profile.
The researchers created an algorithm that focuses Remy's analyses on scenarios in which small network condition variations lead to large performance variations, and to concentrate less on predictable network behavior. "Traditionally, TCP has relatively simple endpoint rules but complex behavior when you actually use it," says MIT graduate student Keith Winstein. "With Remy, the opposite is true. We think that's better, because computers are good at dealing with complexity. It's the behavior you want to be simple."
From MIT News
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA