Apple hires...

Researchers claim to have developed an inexpensive software tool that can diagnose and correct the thousands of tiny network delays responsible for crippling parallel processing speeds in computer clusters.

Researchers at the University of California say their "Lossy Difference Aggregator" could be integrated into today's routers at almost no financial or performance cost and diagnose delays down to tens of microseconds and packet losses that are as infrequent as one in a million.

According to the scientists, if an investment bank's trading algorithm reacts to information from an incoming market data feed just 100 microseconds earlier than the competition, it can buy millions of cheap shares and bid up the stock before its competitors' programs can react.

"Routers today aren't capable of tracking delays through them at microsecond time scales, so exchanges such as the London Stock Exchange use specially crafted external boxes to track delays at various key points in the data center network," says Alex Snoeren, a computer science professor on the project.

But these external systems are too expensive to be added to every router in a data center network, which makes it difficult for network managers to identify and locate flaky routers.

Increasingly, software - rather than humans - responds to streams of information moving across networks in real time and when that happens millionths of seconds count.

“When it comes to fault isolation, networks are a big black box. You put packets in on one side and you get them out the other side,” says Snoeren. “A lightweight network monitoring approach such as ours allows you to pinpoint the source of the performance degradation and identify the problem routers.

This is diagnostic tool, a potentially extremely important one. You don’t want to just know that you have a network problem, you want to know which router and which application is causing the problem.

A typical router may process 50 million packets a second, and keeping track of each packet’s arrival and departure with conventional methods is a daunting piece of bookkeeping, which becomes impoossible if any of the packets have been lost in transit.