Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Cisco Systems, Inc. (CSCO) -- Ignore unavailable to you. Want to Upgrade?

To: JRH who wrote (13485)	4/17/1998 12:09:00 PM
From: Chris	Respond to of 77400

Researchers work to eliminate Net bottlenecks By James Glave SAN FRANCISCO (Wired) - What's a few hundred nanoseconds between friends? An eternity, really, when you realize that's all the time today's Internet routers have to look at a speeding packet, figure out where it's headed, and send it on its way before it gets rear-ended by the next one in line. Recognizing that the biggest bottlenecks lie in routers, George Varghese and his team at Washington University in Saint Louis have come up with two distinct solutions to reduce the time needed to look up a message's address prefix and fire the message back out into the ether. Routers need to know more than 40,000 prefixes, but the length of those prefixes varies from 8 to 32 bits. For example, there's a database for all the 25-bit prefixes, one for 26 bits, et cetera. Varghese's schemes could cut the average prefix lookup time from 1.2 microseconds to 100 nanoseconds (a factor of 10). Varghese's first and simplest method transforms a router database containing 32 possible distinct prefix lengths into one containing a much smaller number, with the help of a kind of binary wild-card scheme. "If we think of prefixes as eggs and prefix lengths as baskets, we are essentially increasing the number of eggs but putting those eggs into fewer baskets," he says. The second idea, binary search on prefix lengths, relies on an algorithm built on a binary tree model. Like a game of Twenty Questions, a yes or no response from the database halves the remaining prefixes until the correct address is discovered. "We can handle the current Internet with live questions," Varghese explains. But what about the Net five years from now, which will have to support 128-bit prefixes due to the increase in the number of addresses (your shoes will eventually need one) and the rise in traffic (blame videostreaming)? The second solution scales for such growth nicely. Varghese has licensed the algorithms to two major router manufacturers, a third and a fourth deal are in the works, and his solutions should be built into the plumbing of the Net long before the arrival of 128-bit addressing causes any bad traffic accidents. dailynews.yahoo.com