A decade ago, when John Spinelli was doing research on the prospects of major computer network failure, no one could relate.
“Now that everyone uses computers, they understand the effects of a network failure,” said the associate professor of electrical engineering and computer science. Spinelli is to give a faculty colloquium titled “Will the Internet Crash: The Reliability of Large Computer Networks” on Tuesday, May 15, at 11:30 a.m. in Reamer Auditorium.
“I am not talking about a few local events, but the kind of structural problems that, though exceedingly unlikely, would be equivalent to an electrical blackout in the entire Northeast U.S.,” Spinelli said.
“Fixing that (type of widespread failure) is not as simple as throwing a switch,” he said.
The Internet is designed as a decentralized or distributed network, Spinelli explains. But having no single point of authority leads to interdependencies that can cause a type of “cascading failure.” Though highly unlikely, conditions that can produce those types of failures are worth studying from an engineering standpoint, Spinelli says.
So, how do you design a distributed system so that keeps local failure local? The answer, Spinelli says, is in a field known as “self-stabilizing systems.” Most network problems begin with a router acting outside a standard mode of operation, where its actions (and its effects on other systems) become unpredictable. Self-stabilizing systems have the ability to recover from starting off in an arbitrary state. While such systems are recognized as good design models, they have not yet been implemented in many distributed networks, he says.
There is something of a Catch-22 with the Internet: a distributed network relies on cooperation. But cooperation also makes the network susceptible to more problems. “To have cooperation, you have to have confidence that the system is reliable,” Spinelli says.
Spinelli, a member of the Union faculty since 1989, holds a bachelor's degree from the Cooper Union, and master's and Ph.D. degrees from MIT.
Spinelli says he is not surprised by the technological improvements in computer networks over the last decade. He is surprised, however, by the extent to which the public has embraced the technology.