Autonomic Computing: Panacea or Poppycock

2005·0.48 MB·English

by  Sterritt, Roy

-
Autonomic Computing Panacea or Poppycock?
EASe 2005 Birds of a Feather Session
Roy Stemtt Mike Hinchey
School of Computing and Mathematics NASA Goddard Space Flight Center
Faculv of Engineering Sofiare Engineering Laboratory
University of Ulster Greenbelt, MD 20771
Northern Ireland USA
rstemtt @ulster.ac.uk michae1.g.hinchey @nasa.gov
pama-ce-a (pha see a)
1. Defiitions' n.
A remedy for all diseases, evils, or difficulties;
a cure-all.
awto-nom-ic (hta n6mmik)
[Latin panacea, from Greek panakeia, from panakes, all-
adj. healing : pan-, pan- + &OS, cure.]
1. Phvsiolom
a. Of,r elating to, or controlled by the
autonomic nervous system.
b. Occurring involuntarily; automatic: an
autonomic rejkx.
2. Resulting from internal stimuli; spontaneous.
[Dutch dialectal pqpekak : pap, pap (from Middle
Dutch pappe, perhaps from Latin pappa,food) + kak,
au-ton.o-&i-ty (gwta n6m i sittee)
dung (from M e n ,t o defecate, from Middle Dutch
n.
kacken, from Latin cacare.]
1. The state of being autonomic.
au-tonwmous (aw t6nnamas)
2. A Brief History of Autonomicity
adj.
1. Not controlled by others or by outside forces;
independent: an autonomous judiciary; an Autonomic Computing arose out of a need for a
autonomous division of a corporate means to cope with rapidly growing complexity of
conglomerate. integrating, managing, and operating computer-based
2. Independent in mind or judgment; self-directed. systems as well as a need to reduce the total cost of
3. ownership of today's systems.
a. Independent of the laws of another Autonomic Computing (AC) as a discipline was
state or government; self-governing. proposed by IBM in 2001, with the vision to develop
b. Of or relating to a Self-governing self-managing systems [l]. As the name implies, the
entity: an autonomous legishare. influence for the new paradigm is the human body's
c. Self-governingw ith respect to local or autonomic system, which regulates vital bodily functions
internal affairs: an autorw-mous region such as the control of heart rate, the body's temperature
of a cowry. and blood flow-all without conscious effort.
4. Autonomic. The vision is to create selfivare through self-*
[From Greek autonomos : auto-, auto- + nomos, law] properties. The initial set of properties, in terms of
objectives, were self-configuring, self-healing, self-
optimizing and self-protecting, along with attributes of
self-awareness, self--monitoring and self-adjusting. This
self-* list has grown: self-anticipating, self-critical, self-
defining, self-destructing, self-diagnosis, self-governing,
Adapted from American Heritage Dictionary of the
English Language, 4' edition.
self-organized, self-reflecting, and self-simulation, for (DATC), in order to benefit from the
instance [2][3]. methodologies of control theory, and
simultaneously address the challenges of applying
control theory to computing systems.
3. Some EASe Success Stories Randles et ai. [ 141 highlight the need for a meta-
framework for self-governance, in order to
Although the Autonomic Computing initiative itself is achieve large-scale autonomicity.
still very much in its infancy, there have been, Wang and Mathur [l5] present an interceptor-
nevertheless, a number of success stories in developing based approach for constraint-violationd etection,
autonomic systems reported at EASe 2004 and EASe for which monitor code can be generated
2005, along with the identification of a number of automatically from XML-based constraint
challenges: specifications.
0 The potential benefit of using well-established
Rouff et al. [4] [8]d escribe a forthcoming NASA systems engineering concepts and techniques in
mission which is a concept for future missions the development of complex systems is evaluated
involving swarm technology, which will by Bustard et al. [ 161, with particular reference to
necessitate autonomic behavior, and which two existing well-established methodologies.
highlights great challenges for verification of this An autonomic system that supports transparent
class of systems. stream synchronization in multimedia streaming
Sterritt et al. [5] highlight the importance of a applications, with necessary components installed
reflex-healing dual strategy to facilitate the on the fly, is described by Friedland and Pauls
addition of autonomic capabilities to the ~171.
telecommunications fault management An autonomic system integration platform where
architecture. holistic design models capture system structure
Nichols and Bapty [6] describe an adaptive image and target system resources and autonomic
processing environment that allows solutions of behavior, is proposed by Shetty et al. [18].
complex image processing problems to be built Rash et af. [19] describe a tool to support fully
and executed rapidly on a number of hardware formal requirements-based programming, and
architectures. describe its application to a system that exhibits
Shew et al. [7] describe a language to define autonomic properties.
behaviors for the BTeV trigger system, which is A comprehensive prototype autonomic system, in
being used as a model for tools for defining fault which self-optimizing agents provide a self-
behavior and automatically generating software. healing layer with ability to discover, diagnose,
Tmzkowski et ai. 191 describe the autonomic and react KO discontinuities in real-time
properties of two NASA legacy multi-agent processing, is discussed by Messie et al. [20].
systems, namely Agent Concept Testbed (ACT) Sterritt and Hinchey [21] describe apoptosis and
and the Lights-Out Ground Operations System self-destruction, the ‘‘ultimate” self-protection
(LOGOS). mechanism in autonomic systems, as applied to
Gracanin et ai. [lo] describe the use of a model- space missions.
based architecture for the development of PACT - Personal Autonomic Computing Tools -
autonomic systems, using the COUGAAR describes efforts, in terms of prototypes, to
architecture as a platform. increase the degree of autonomicity in today’s
Caseau [ 1I ] describes a set of adaptive methods personal systems [22].
and rules for routing messages in an Enterprise Baldassari et al. [23] describe an experimental
Application Environment, which yields a form of cluster management system, which, although
autonomic behavior. small, demonstrates a decrease in overhead as the
Stemtt and Chung [12] describe a proof-of- cluster size grows.
concept self-healing tool for the personal A theoretical protocol for autonomic distribution
computing environment, which incorporates a of services in a P2P environment is presented by
pulse monitor and vital signs health monitor. Saffre and Blok [24], who aim to demonstrate
The correspondence between elements of control that distribution that meets the requirements of
systems and those of autonomic systems is the community can be achieved without
identified by Diao et al. 1131, who propose a centralized resource management.
Deployable Testbed for Autonomic Computing
4. Challenges for the Future Lighthill Report, which all but ended funding for AI
research in the UK at the time.
4.1 Learning from Past Experience
43 A Balanced View
Several promising fields of engineering and computer
We believe that Autonomic Computing has much to
science have suffered badly ftom unwarranted claim,
offer in the advancement of complex computer-based
exaggeration of benefits, or extrapolation of minor
systems.
results to make claims of finding the “holy &.”
We expect to see many additional self- properties
Formal methods for system specification and design,
being added to the portfolio of behaviors expected of an
for example, have failed to live up to the exaggerated
autonomic computing system. We anticipate many new
claims that were made about their benefits. Unfounded
biologically-inspired metaphors being developed and
claims included that formal methods would result in
incorporated into future autonomic system.
fully correct computer systems [25]. What protagonists
We believe, however, that the community needs to
failed to consider is that they would result in systems
keep a balanced view. While we see autonomic
that were correct with respect to their specifications. If
computing and the autonomic metaphor as being a major
specifications were flawed then too would be the
step forward and a useful contribution to the future of
resulting systems. Of course, formal methods offer
Computing, we must be careful not to make unfounded
opportunities to uncover errors in specifications,
claims for it, nor to have unreasonable expectations of it.
resulting in high quality software systems that are often
cheaper to develop since errors are uncovered early
43 Panacea of Poppycock?
[26][27]. Notwithstanding these benefits, formal
methods are often considered to have failed to meet up
We certainly don’t see Autonomic Computing as
to the claims made for them. Unfortunately, these
“poppycock”, but we realize we’re probably preaching
claims were unreasonable.
to the choir on that.
Neural networks, evolutionary programming, and
Nor, however, do we see it as a panacea. In fact, we
fuzzy logic have all suffered similar fates. Despite
do not believe that any such panacea exists nor could
useful results and cxmtributions to the advancement of
exist. The development of computer-based systems
Computing, early claims that these approaches could
involves the complex interactions between a number of
offer some form of ‘‘magtc wand” that would solve a
stakeholders, meeting a variety of desiderata such as
multitude of problems, or even all problems, simply
cost, lead-time, functional and non-functional
could not be lived up to. The result is a cynical view by
requirements, development standards both self-imposed
the public, funding agencies, and even researchers
(e.g., quality standards, or for certification purposes)
themselves, as to the benefits of these disciplines.
and mandated (by government and other bodies), ease of
A similar situation arose with the field of Artificial
use, maintainability,e tc.
Intelligence itself. In the 1950s and 196Os, protagonists
The Autonomic Computing community is performing
claimed (falsely) to have invented computers that could
some excellent work, both research and applied. We
think. In reality, all they had accomplished at the time
believe that many of the results currently being produced
was a computer program that could solve a single simple
will become standard practice in a variety of domains in
problem. The techniques that were produced were well
the future. This is particularly true in situations where
worth pursuing further, but were far from embodying the
computer systems must be autonomous, either because
elusive “machines that think.”
they are required to perform tasks that are not possible
Herb Simon, in his memoirs 1281, recounts that his
for humans to perform (e.g.. deep space exploration), or
students claim he told them in class one Monday
tasks that require decisions to be made so quickly that
morning that over the weekend he and Alan Newel1 had
waiting for human intervention is just not feasible. We
invented a computer program that could think. Simon
believe that these classes of autonomous systems will
admits that to have made such a claim would be
exhibit autonomic properties more and more in the
ludicrous, and didn’t recall doing so; he concedes,
future. In fact, we believe that the autonomous systems
however, that there were so many “witnesses” that he
of the future will greatly rely on autonomic properties
must have made such a claim.
for their viability, and their very existence.
Such outlandish claims were used by the opponents
We must be careful, however, not to make the
of Artificial Intelligence to “demonstrate” that
mistakes that many other communities have made. That
computers would never be able to do the sorts of things
is, we must be conscious of the limitations of Autonomic
that the proponents of AI were claiming [29], and indeed
Computing, and not make claims for it that we cannot
had a major role in influencing the findings of the 1973
substantiate. However, we must continue to pursue the
field and ensure that its contributions are both developed llth Annual IEEE International Conference and
further and applied in practice. We must demonstrate Workshop on the Engineering of Computer Based
successes and the sustainability of the effort, without Systems (ECBS 2004), Bmo, Czech Republic, 24-27
Mzy 2w, pp 978-483.
succumbing IO the iryperboie inat has stifled progress in
181 W. Truszkowski, J. Rash, C. Rouff, M. Hinchey,
other disciplines.
“Asteroid Exploration with Autonomic Systems”,
Proceedings of IEEE Workshop on the Engineering of
Right now it’s only a notion but I think I can get Autonomic Systems (EASe 2004) at 11th Annual IEEE
money to make it into a concept and then later Iruernational Conference and Workshop on the
change it into an idea. Engineering of Computer Based Systems (ECBS
Woody Allen, “Annie Hail” 2004), Bmo, Czech Republic, 24-27 May 2004, pp
484-489.
Acknowledgements r91 W. Tmszkowski, J. Rash, C. Rouff, M. Hinchey,
“Some Autonomic Properties of Two Legacy Multi-
Agent Systems - LOGOS and ACT”, Proceedings of
A great debt is due to EASe contributors, EASe IEEE Workshop on the Engineering of Autonomic
program committee and TC-ECBS. Systems (EASe 2004) at llth Annual IEEE
International Conference and Workshop on the
References Engineering of Computer Based Systems (ECBS
2004), Bmo, Czech Republic, 24-27 May 2004, pp
490-498.
Ill A.G. Ganek, T.A. Corbi, “The dawning of the
autonomic computing era”, IBM Systems Journal, 1101 D. Gracanin, S.A. Bohner, M. Hinchey, ‘Towards a
Model-Driven Architecture for Autonomic Systems”,
42(1), 2003.
Proceedings of IEEE Workshop on the Engineering of
121 H. Tianfield, Multi-agent based autonomic
Autonomic Systems (EASe 2004) at I1t h Annual IEEE
architecture for network management”, Proceedings of
INDIN 2003, IEEE International Conference on International Conference and Workshop on the
Idustrial Informatics, 21-24 Aug. 2003, pp 462 - Engineering of Cornpurer Based Systems (ECBS
20041, Bmo, Czech Republic, 24-27 May 2004, pp
469.
500-505.
131 R. Stemn, “Autonomic Computing”, Innovations in
1111 Y. Caseau, “Self-Adaptive and Self-Healing Message
Systems and Software Engineering: a NASA Joud,
Passing Strategies for Process-Oriented Integration
Springer, 1(1), April 2005.
Infrastructures”, Proceedings of IEEE Workshop on
141 C. Rouff, A. Vanderbilt, M. Hinchey, W.
the Engineering of Autonomic Systems (EA& 2004) at
Truszkowski, J. Rash, “Verification of Emergent
I lth Annual IEEE International Conference and
Behaviors in Swarm-based Systems”, Proceedings of
Workshop on the Engineering of Computer Based
IEEE Workshop on the Engineering of Autonomic
Systems (ECBS 20041, Bmo, Czech Republic. 24-27
Systems (EASe 2004) at 11th Annual IEEE
May 2004, pp 506-5 12.
International Conference and Workshop on the
[E] Sterritr R, Chung S, “Personal Autonomic Computing
Engineering of Computer Eased Systems (ECBS
Self-Healing Tool”,P roceedings of IEEE Workshop
2004, Bmo, Czech Republic, 24-27 May 2004, pp
on the Engineering of Autonomic Systems (EASe
443-448.
2004) at 11th Annual IEEE Intemtional Conference
[51 R. Stemtt, D. Gunning, A. Meban, P. Henning,
and Workshop on the Engineering of Computer Based
“Exploring Autonomic Options in an Unified Fault
Systems (ECBS 20041, Bmo, Czech Republic, 24-27
Management Architecture through Reflex Reactions
May 2004, pp 513-520.
via Pulse Monitoring”, Proceedings of IEEE
1131 Y. Diao, J.L. Hellerstein, S. Parekh, G. Kaiser, D.
Workshop on the Engineering of Autonomic Systems
Phung, “Self-Managing Systems: A Control Theory
(EASe 2004) at the 11th Annual IEEE International
Foundation”, Proceedings of IEEE Workshop on the
Conference and Workshop on the Engineering of
Engineering of Autonomic Systems (Use 2005) at
Computer Based Sysems (ECBS 2004), Bmo, Czech
12th Annual IEEE International Conference and
Republic, 24-27 May 2004, pp 449-455.
Workshop on the Engineering of Computer Based
[61 J. Nichols, T. Bapty, “A model-based self-adaptive
Systems (ECBS 2005), Greenbelt, MD, USA, 3-8
approach to image processing”, Proceedings of IEEE
April, 2005, pp XXX-XXX.
Workshop on the Engineering of Autonomic Systems
I141 M. Randles, A. Taleb-Bendiab, P. Miseldine, A. Laws,
(EA& 2004) at llth Amuul IEEE International
“Adjustable Deliberation of Self-Managing Systems”,
Conference and Workshop on the Engineering of
Proceedings of IEEE Workshop on the Engineering of
Computer Based Systems (ECBS 2004), Bmo, Czech
Autonomic Systems (EASe 2005) at 12th Annual IEEE
Republic, 24-27 May 2004, pp 456-461.
International Conference and Workshop on the
[71 S. Shetty, S. Neema, T. Bapty, “Model Based Self
Engineering of Computer Based Systems (ECBS
Adaptlve Behavior Language for Large Scale Real-
2005), Greenbelt, MD, USA, 3-8 April, 2005, pp
time Systems”, Proceedings of IEEE Workshop on the xxx-xxx.
Engineering of Autonomic Systems (Use 2004) at
[I51 Q. Wang, A. Mathur, “Interceptor based Constraint R. Stemtt, M.G. Hinchey, “Engineering Ultimate Self-
Violation Detection”, Proceedings of IEEE Workshop Protection in Autonomic Agents for Space Exploration
on the Engineering of Autonomic System (EASe Missions”, Proceedings of IEEE Workshop on the
2@5j ui i2ih Arm-’ 1EEE liizcrmrioid Conference Ezgirie~nso f Au:ono~ic Sys:mii {E?Se 2x5: A*
and Workshop on the Engineering of Computer Based 12th Annual IEEE International Conference and
Systems (ECBS 2005). Greenbelt, MD, us& 3-8 Workshop on the Engineering of Computer Based
April, 2005, pp XXX-XXX. Systems (ECBS 2005), Greenbelt, MD, USA, 3-8
[I61 D. Bustard, R. Stemtt, A. Taleb-Bendiab, A. Laws, M. April, 2005, pp XXX-XXX.
Randles, F. Keenan ‘Towards a Systemic Approach to t 221 R. Stemtt, TACT: Personal Autonomic Computing
Autonomic Systems Engineering”, Proceedings of Tools”, Proceedings of IEEE Workshop on the
IEEE Workshop on the Engineering of Autonomic Engineering of Autonomic System (EASe 2m5) QI
System (EASe 2005) at 12th A d IE EE 12th A n d IEEE Intcnrational Conference and
International Conference and Workshop on the Workshop on the Engineering of Computer Based
Engineering of Computer Based System (ECBS Systems (ECBS 2003, Greenbelt, MD, USA, 3-8
2005), Greenbelt, MD, USA, 3-8 Apnl, 2005, pp April, 2005, p~ XXX-XXX.
xxx-xxx. 1231 J. D. Baldassari, C. L. Kopec, E. S. Leshay, W.
[ 171 G. Friedland, K. Pauls, “Towards a Demand Driven, Truszkowski, D. Finkel, “Autonomic Cluster
Autonomous Processing and Streaming Architecture”, Management System (ACMS): A Demonstration of
Proceedings of IEEE Workshop on the Engineering of Autonomic Principles at Work”, Proceedings of IEEE
Autonomic Systems (EASe 2005) at 12th Annual IEEE Workshop on the Engineering of Autonomic Systems
International Conference and Workshop on the (EASe 2005) at 12th Annual IEEE International
Engineering of Computer Based System (ECBS Conference and Workshop on the Engineering of
2005), Greenbelt, MD, USA, 3-8 April, 2005, pp Computer Based Systems (ECBS 2005), Greenbelt,
xxx-xxx. MD. USA, 3-8 April, 2005, pp XXX-XXX.
[IS] S. Shetty, S. Nordstrom, S. Ahuja, D. Yao, T. Bqty, F. Saffre, H. R. Blok “Selfservice: Self-organised
S. Neema, “Systems Integration of Large Scale distribution of services for P2P communities”,
Autonomic Systems using Multiple Domain Specific Proceedings of IEEE Workshop on the Engineering of
Modeling Language”, Proceedings of IEEE Workshop Autonomic Systems (EASe 2005) at 12th Annual IEEE
on the Engineering of Autonomk System (EASe International Conference and Workshop on the
2005) at 12th Annual IEEE International Conference Engineering of Computer Based Systems (ECBS
and Workshop on the Engineering of Computer Based 2003, Greenbelt, MD, USA, 3-8 April, 2005, pp
System (ECBS 200.5). Greenbelt, MD, USA, 3-8 xxx-XXX.
April, 2005, pp XXX-XXX. J.P. Bowen and M.G. Hinchey, “’Tis Greek to Me:
[19] J. Rash, M. Hinchey, C. Rouff, D. W i n , Method in the Madness”. In M.G. Hinchey and J.P.
“Experiences with a Requirements-Based Bowen, editors, Industrial Strength Formal Methodr
Programming Approach to the Development of a in Practice, Springer Verlag FACIT series, London,
NASA Autonomous Ground Control System”, 1999.
Pioceedings oflEEE Forkshop OR :k Erigineeiing of M.G. Hinchey and I.P. Bowen, “Applications of
Autonomic Systems (EASe 2005) at 12th Annual IEEE Formal Methods FAQ’. In M.G. Hinchey and J.P.
International Conference and Workshop on the Bowen, editors, Applications of Formal Methods,
Engineering of Computer Based Systems (ECBS Prentice Hall, Heme1 Hempstead and Englewood
2005), Greenbelt, MD, USA, 3-8 April, 2005, pp Cliffs, 1995.
xxx-xxx. J.P. Bowen and M.G. Hinchey, Seven More Myths of
1201 D. Messie, M. Jung, J. C. Oh, “Prototype of Fault Formal Methods. IEEE Software, 12(4):3441, July
Adaptive Embedded Software for Large-Scale Real- 1995.
Time Systems”, Proceedings of IEEE Workshop on 1281 H.A. Simon, Modeh of My Life, Basic Books, NY,
the Engineering of Autonomic Systems ( M e2 005) at 1991.
12th Annual IEEE International Conference ana’ 1291 H.L. Dreyfk, Whar Computers Still Can’t Do: A
Workshop on the Engineering of Computer Based Critique of Artificial Reason, MIT Press, Cambridge,
Systems (ECBS 2005), Greenbelt, MD, USA, 3-8 MA, 1992.
April, 200.5,p p XXX-XXX.