04 Future Directions For

1
Future Directions for Autonomous Decentralized Systems

(Opening Statement of the Panel)
H. Kopetz
Technische Universitt Wien, Austria
hk@vmars.tuwien.ac.at
To me, the future for autonomous decentralized
systems (ADS) looks bright, because of ongoing
developments in the following four areas
(i)
system complexity
(ii)
hardware architecture
(iii) legacy systems
(iv) fault-tolerance.
Let me discuss some of the issues in each one of
these areas in detail.
System
Complexity:
The conceptual
management of the every increasing complexity of large
computer-based systems is, in my opinion, one of the
main current and future challenges of our profession. The
"unmanaged system complexity" is often the major cause
for schedule overruns, missing dependability and often
outright project failure (see, e.g., the very revealing
report on the US Advanced Automation System (GAO
1997). One established technique to handle this
complexity is the introduction of a rigid system structure
that partitions a large systems into nearly autonomous
subsystems and that reduces the interactions among these
subsystems to the minimum information flow required for
the achievement of the system objectives (Kopetz 1997,
p.272). The stable interfaces between these subsystems
must be fully defined in the temporal domain and in the
value domain, carefully controlled, and above all,
understandable in their structure and behavior. The
interfaces must also act as error propagation boundaries.
From the point of view of subsystem A that interfaces to
a subsystem B, the shared interface definition must
contain all information that is necessary for subsystem A
to perform its mission (and vice versa). With respect to
subsystem B, the proper operation of subsystem A thus
depends only on the correct and timely information in the
interface. The interface specifications thus states the
mutual obligations of the cooperating subsystems. From
the point of view of subsystem B, every implementation

of subsystem A that meets this established interface
specification is adequate. In such an architecture
subsystem A can act autonomously as long as it meets
the constraints of the interfaces. From the vantage point
of a higher level, the system operation can be understood
by grasping the information flow across the interfaces,
disregarding the internal operation of the autonomously
operating subsystems.
The simplicity and
understandability of the subsystem interfaces is thus the
key to mastering the complexity at the system level.
Hardware Architecture: The most cost-effective
structure for a large computer system depends on the
optimal combination of hardware and software resources.
Because of the phenomenal advances of the
microelectronics technology, the hardware part of the cost
of implementing a large system is shrinking to the point
where it makes not only technical, but also economic
sense to design system architectures that are guided by
clean functional partitioning according to the established
architectural principle "form follows functions". The
systems on chip (SOC) that are available today--a
powerful microprocessor with sufficient memory, I/O
interfaces and a communication controller to implement a
self-contained application function--enable the design of
hardware structures that do not make any compromises for
the sake of saving on hardware resources. A good
example of this kind of thinking is the use of a system
chip with a Power PC processor for the control of airbags
in a automobile. Although such a system chip might only
be used once in its lifetime for a critical time interval of a
few milliseconds--the point in time between detecting an
impact and the point in time of activating the airbags--it
makes architecturally and economically sense to assign a
dedicated piece of hardware to this task, not being detracted
by resource utilization arguments. In the future I see these
SOCs with the appropriate application software as
forming autonomous subsystems of large computer

systems. Again, the most important research issue is the
simplicity and understandability of the subsystem
interfaces.
Legacy Systems: Many of the future computer
applications will not be designed on the "green lawn", but
will be extensions and interconnections of already existing
computer systems. These "legacy system", developed
years ago as stand-alone operations, must be integrated
into new distributed architectures with minimal
interruptions of the day-to-day operations. The most
promising technology to achieve these objectives is the
"wrapper" technology: the provision of a gateway
component that on the one side adapts to the idiosyncratic
interface of the legacy system and on the other side
conforms to the standard interface specification of the new
distributed architecture. From the architecture point of
view, the legacy system together with the associated
gateway component, can be seen as an autonomous
decentralized systems that provides the required
functionality across the standardized interface of the
gateway component.
Fault Tolerance: It is a law of physics that every
physical component is going to fail eventually. Although
the failure rate of a single VLSI chips is low (between
10 -7 and 10-8 failures/hours for permanent physical
failures), the failure rate of a non-fault-tolerant large
system that comprise many chips can be significant.

Considering the low cost of the computer hardware, it is
becoming cost-effective to reduce this system failure rate
by the systematic provision of fault tolerance in many
safety-critical or money-critical applications. A
prerequisite for the successful application of fault tolerance
is the provision of a system architecture that supports the
introduction of autonomous error-containment regions and
fault-containment regions with error propagation
boundaries around them. A large system, consisting of
many decentralized autonomous subsystems will small
understandable interfaces around them, comes close to this
primary requirement on a fault-tolerant architecture.
References:
GAO (1997). Air Traffic Control--Complete and Enforced
Architecture Neded for FAA Systems
Modernization. US General Accounting Office
(GAO).
Kopetz, H. (1997). Real-Time Systems, Design

Principles for Distributed Embedded Applications;
ISBN: 0-7923-9894-7. Boston. Kluwer Academic
Publishers.

04 Future Directions For

Uploaded by

Copyright:

Available Formats

04 Future Directions For

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

04 Future Directions For

Uploaded by

Copyright:

Available Formats

1

Future Directions for Autonomous Decentralized Systems

the point of view of subsystem B, every implementation

forming autonomous subsystems of large computer

system that comprise many chips can be significant.

Kopetz, H. (1997). Real-Time Systems, Design

You might also like