Selective shutdown protects nets

By Kimberly Patch, Technology Research News

Networks, including the Internet and power grids, are complicated entities. They are made up of nodes -- servers or power stations -- that connect to greater or lesser degrees with other nodes, and they are susceptible to cascade failures, which occur when large, well-connected nodes abruptly disconnect from the network.

A researcher from the Max Planck Institute for the Physics of Complex Systems in Germany has shown that it might be possible to suppress cascade failures triggered by attacks on large network nodes by shutting down peripheral nodes, much like a forest fire can be controlled by setting small containment fires.

The strategy could eventually be used to reduce the damage of attacks on key network nodes. The work also advances the understanding of the interplay between network structure and network dynamics, said Adilson Motter, a guest scientist at the Max Planck Institute.

Previous research has shown that eliminating large, central network nodes like the backbone routers of the Internet or large power stations can generate a cascade of failures of other nodes, but eliminating smaller peripheral nodes is likely to have little effect on the network as a whole.

Motter constructed a model that shows that the size of a cascade can be drastically reduced if a certain number of nodes that handle small loads are taken out of service before the cascade effect begins. "When central nodes are attacked, the intentional elimination of peripheral nodes right after the initial attack, before the propagation of the cascade, may be the key to [reducing] the size of an otherwise large cascade," said Motter.

Key to the method is identifying the right nodes to shut down, said Motter. "The main challenge was to find the correct nodes and/or lines to be removed in the control of the cascade," he said. The wrong choices can enhance the size of the cascade rather than reducing it, he added.

The idea is to keep the overall network load as balanced as possible. Nodes have two simultaneous functions. They act as transmitters, which distribute load, and generators, which create load. Central nodes do more transmitting than generating, which makes them important in load balancing but also makes them targets for attacks that would disrupt the network.

Removing nodes that generate significantly more than they transmit reduces the overall load. The model shows that removing load-generating nodes can dampen cascade failures generated by abruptly shifting loads.

The method also works if the defensive strategy calls for shutting down heavily-loaded links, which carry traffic from load-generating nodes to central distribution nodes, rather than targeting load-generating nodes.

The model implies that the method could be used to reduce the effects of major failures in networks like the Internet and power stations, but the model is simple enough that it doesn't account for all the features of specific systems, said Motter. "It is still speculative to talk about practical applications [but] I hope to my work will motivate new studies on the control of cascading failures in realistic models of network systems," he said.

To illustrate the idea, however, consider power grids, said Motter. Power grids have three different types of power stations: generators, which are the power sources, local stations that distribute power to customers, and transmission stations the transfer powers from generators to local stations.

The propagation of cascading failures in power grids is determined by automatic devices along the transmission lines that take components of an a grid out of service when the load on them becomes too high, said Motter.

"These devices are designed to protect the power grids against permanent damage, but may eventually disconnect a substantial part of the network," he said.

The stations most likely to be directly affected by the cascade are transmission stations, and the overload failure of one transmission station may disconnect several local stations and/or cause overloads in other transmission stations, said Motter. "Before the propagation of the cascade there's enough time, in principle, to switch off suitable power stations in order to reduce the load on the grid in a controlled way."

Motter's model suggests that the size of the cascade can be reduced if local stations are intentionally disconnected from the transmission stations that are about to fail. "More work is needed in order to have realistic models that take all the details of a network into account," Motter said. "This is a topic of my current research," he added.

The research was funded by the Max Planck Institute for the Physics of Complex Systems.

Timeline:   Unknown
Funding:   Institute
TRN Categories:  Physics; Networking; Internet
Story Type:   News
Related Elements:  Technical paper, "Cascade Control in Complex Networks," posted at the arxiv physics archive at


August 25/September 1, 2004

Page One

Selective shutdown protects nets

Tools design DNA-nanotube logic

Five photons linked

Liquid crystal IDs pathogens

Photosynthesis drives solar cell
Hybrid nanowire makes transistor
Nanocrystals spark efficient LEDs
Nanotubes make fluid filter
DNA copier uses little power
Method makes stronger steel


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.