8 1541-1672/06/$20.00 © 2006 IEEE IEEE INTELLIGENT SYSTEMSPublished by the IEEE Computer Society

G u e s t E d i t o r s ’ I n t r o d u c t i o n

Self-Management throughSelf-Organization

Márk Jelasity and Ozalp Babaoglu, Università di Bologna

Robert Laddaga, Massachusetts Institute of Technology

G iven the scale and complexity of today’s information systems, it’s increasingly

important that they handle system management problems and tasks themselves—

intelligently and autonomously. They must be able to work around errors and failures

as well as adapt their configuration for optimal response to actual system state or

environment—with minimal or no human interven-tion. Self-management can significantly reduce costsand increase robustness. Moreover, it’s sometimesrequired because time or location constraints makehuman management impossible. Bringing self-management about in a variety of systems such aswired or wireless networks, peer-to-peer systems,the computational Grid, and distributed and embed-ded systems and applications in general is an excit-ing and quickly growing research area.

This special issue focuses on one particularapproach, motivated by observations of the self-organizing and self-managing systems that surroundus—multicellular organisms, social insects, marketeconomies, human societies, ecosystems, and so on.These systems are made of components that obeysome local rules and act on the basis of local obser-vations—often selfishly. Yet the system as a wholeexhibits global properties such as self-healing, self-tuning, and self-organization—exactly the proper-ties we’re after for complex information systems.Distilling the key ideas from these systems and incor-porating them into information systems often leadsto inexpensive, straightforward, and highly robustsolutions.

Self-organizationIt’s difficult to define self-organization in a way

that satisfies everyone. Perhaps a definition isn’t soimportant or productive. Instead, we can work withan intuitive concept that emphasizes decentraliza-tion and emergent functionality. In this sense, a self-organizing system consists of many interacting com-

ponents that have partial or no global systemknowledge. The components interact locallywith other system components according tosome network of possible interactions. Thecomponents are autonomous in that no cen-tral controller directs their actions towardglobal goals according to current system stateinformation. Instead, the components deter-mine their own actions, taking as input onlylocal information available from their neigh-bors in the communication topology and theenvironment. Despite this local behavior, thesystem as a whole meets its global goals.

In most cases, the components don’t eveninclude representations of the system’s globalstate and global goals. They might not evenhave individual goals. In other cases, thecomponents might know the global goalscompletely or partially but not know exactlyhow to cooperate to achieve them. In yetother cases, components might have indi-vidual goals that contradict the system goalsin one way or another.

Self-managing systems are unquestion-ably interesting simply because many ofthem are crucial to us—society, a living cell,ecosystems, and so on. These are by defini-tion self-organizing, and we need to under-stand them as best as we can.

But do we really want or need to build self-organizing information systems? After all,controlling a system from a central locationis much easier and more convenient. It’s anatural thing to do, and many times it’s alsothe right thing to do. However, in someimportant cases, self-organization is useful,and at times it’s necessary.

For example, central coordination isn’tfeasible in systems such as sensor networks,mobile ad hoc networks, and other embed-ded or spatial networks. The devices thatform the network simply aren’t powerfulenough for their radio signals to reach a com-mon central location, and relay stationsmight not be available.

Nor is central control feasible when systemcomponents belong to different administrativedomains that aren’t willing to cede control.Prime examples of this are the Internet itselfand, more recently, peer-to-peer networks.

Even when not required, a self-organizingdecentralized design might be useful. Forexample, if reliability is a top system prior-ity, you want to avoid any single points offailure, such as a controller. The traditionalway of dealing with this is to replicate func-tionality, say with backup controllers; but atruly self-organizing design can achieve sig-

nificantly higher reliability levels with nocritical system elements. Or when performanceis important, self-organizing decentralizeddesigns can offer greater parallelism and avoidcommunication bottlenecks more easily.

In this issueThis issue includes five articles that rep-

resent some of the most important techniquesand application areas of self-organization.

“Infrastructure for Engineered Emergenceon Sensor/Actuator Networks,” by Jacob Bealand Jonathan Bachrach, discusses the Protolanguage and its implementation for high-level programming of sensor-actuator net-works. Proto helps in designing complex coor-dination and data processing operations overspatial networks of arbitrary size and density.The authors demonstrate the approach by gen-erating synchronized wave patterns, both insimulations and in real hardware.

“Extended Stigmergy in Collective Con-struction,” by Justin Werfel and RadhikaNagpal, presents algorithms that allow forcompletely independent mobile robots to col-laborate in building physical structures. Therobots communicate with each other implic-itly, through the structure itself. The authorscompare and analyze several possibilities forimplementing this stigmergic communication.

“SLACER: A Self-Organizing Protocol forCoordination in Peer-to-Peer Networks,” byDavid Hales and Stefano Arteconi, presents aprotocol that lets a peer-to-peer system opti-mize the sum of the peers’ utility, where indi-vidual utility is a function of mutual interac-tions among the peers. In a Prisoner’s Dilemmagame, the protocol maintains a high coopera-tion level and therefore good overall utility.

“Self-Organization Patterns in Wasp andOpen Source Communities,” by Sergi Valverde,Guy Theraulaz, Jacques Gautrais, VincentFourcassié, and Ricard V. Solé, demonstratesuniversal self-organization principles bydrawing parallels between wasps and opensource developer communities. This work isan excellent reminder that self-organizationis at work in large and complex informationsystems, whether we apply it consciously ornot.

Finally, “Customer-Driven Sensor Man-agement,” by Tracy Mullen, ViswanathAvasarala, and David L. Hall, applies a mar-ket-based approach to allocating sensors totasks. They transform a bidding procedureinto an optimization problem that, whensolved or approximated, results in an opti-mal or good assignment.

To give readers a broader spectrum ofideas from this exciting field, we invited

leading experts from various disciplines toinformally summarize their research positions.Under the general title “InterdisciplinaryResearch: Roles for Self-Organization,” thesefour invited contributions close the issue.

MARCH/APRIL 2006 www.computer.org/intelligent 9

T h e A u t h o r sMárk Jelasity is aresearcher in the De-partment of ComputerScience at the Univer-sity of Bologna, Italy,and a senior researchscientist in the Re-search Group on Arti-ficial Intelligence at the

University of Szeged and Hungarian Academyof Sciences. His research interests include peer-to-peer systems, reliable large-scale distributedsystems and computing, and bioinspired andevolutionary computing. He received his PhDin computer science from the University of Lei-den. Contact him at the Università di Bologna,Dipartimento di Scienze dell’Informazione,Mura Anteo Zamboni 7, I-40126 Bologna, Italy;jelasity@cs.unibo.it.

Ozalp Babaoglu is aprofessor of computerscience at the Univer-sity of Bologna, Italy.His research activitiesinclude distributed com-puting and complexadaptive systems. Hereceived his PhD in

computer science from the University of Cali-fornia at Berkeley, where he was a principaldesigner of BSD Unix. In 1993, he received theUsenix Association Lifetime AchievementAward for his contributions to the Unix systemcommunity and to open industry standards. He’san ACM fellow and serves on the editorialboards for ACM Transactions on Computer Sys-tems, ACM Transactions on Autonomous andAdaptive Systems, and Distributed Computing.Contact him at the Università di Bologna,Dipartimento di Scienze dell’Informazione,Mura Anteo Zamboni 7, I-40126 Bologna, Italy;babaoglu@cs.unibo.it.

Robert Laddaga is aresearch scientist at theMassachusetts Instituteof Technology Com-puter Science and Arti-ficial Intelligence Lab-oratory. His researchactivities include self-adaptive software tech-

nology, perceptually enabled interfaces, soft-ware for physically embedded autonomoussystems, and programming language design. Hereceived his PhD in philosophy of science fromStanford University. Contact him at MITCSAIL, 32 Vassar St., Cambridge, MA 02139;rladdaga@csail.mit.edu.