Global Computing Proactive Initiative:

1
Global Computing Proactive Initiative Cluster
Foundations of Networks and Large Distributed
Environments
Paul Spirakis Computer Technology Institute and
Patras University
IST-FET Workshop on GLOBAL COMPUTING Malaga,
Spain,July 11, 2002
2
OVERVIEW OF THE TALK

• Definition of FET/Definition of GC
• Projects in the Cluster
• Description of each Project
• Partners
• Objectives
• Description of Work
• Expected Results
• Cluster Structure
• Discussion

3
Definition of FET

• The purpose of FET is to promote research that is
  of a longer-term nature or involves particularly
  high risks compensated by the promise of major
  advances and the potential of industrial or
  societal impact.
• It does so by looking with an open mind towards
  the horizon of emerging research opportunities.
• It can be seen as the nursery of novel and
  emergent ideas, some of which may become the
  mainstream topics of the future.

4
Definition of GC

• The research should concentrate on systems having
  the following characteristics
• The systems are composed of autonomous
  computational entities where activity is not
  centrally controlled.
• The computational entities are mobile.
• The configuration varies over time
• The systems operate with incomplete information
  about the environment.
• The ultimate goal of the research action is to
  provide a solid scientific foundation for the
  design of such systems, and to lay the groundwork
  for achieving effective principles for building
  and analysing such systems.

5
PROJECTS IN THE CLUSTER
1) FLAGS Foundational Aspects of Global
Computing Systems
2) CRESCCO Critical Resource Sharing for
Cooperation in Complex Systems
3) DBGLOBE A Data-centric Approach to Global
Computing
4) SOCS A computational logic model for the
description, analysis and verification of global
and open societies of heterogeneous computees
6
FLAGS - Partners
1. Computer Technology Institute (CTI),
leader Prof. Paul Spirakis 2.
University of Athens (UoA),
leader Prof. Elias Koutsoupias 3. University of
Cyprus (UCY), leader Prof. Marios
Mavronicolas 4. University of Paderborn (UPB),
leader Prof. Burkhard Monien 5. Universitat
Politècnica de Catalunya (UPC), leader Prof.
Josep Diaz
7
FLAGS-Objectives
New global computing and communication
environments are emerging that integrate (a)
autonomous, interacting, selfish entities, (b)
highly dynamic multi-agent environments and (c)
ad-hoc mobile networks. For the efficient and
robust implementation of global computing
scenarios in such systems, the project aims to
provide a unifying foundational framework and a
coherent set of design rules, for (i)
co-operation and antagonism of autonomous
entities (ii) stability and fault-tolerance in
multi-agent environments and (iii) communication
and motion in ad-hoc mobile networks.
8
FLAGS-DESCRIPTION OF WORK

• a) co-operation and antagonism of selfish
  entities
• how easy/hard is it to find Nash Equilibria?
• design mechanisms for games
• b) stability
• instability and stability bounds for FIFO
  queuing networks
• heterogeneous networks of mixings of queueing
  policies
• c) wireless networks
• robust and efficient communication in ad-hoc
  mobile networks
• smart-dust sensor protocols for local detection
  and propagation

9
FLAGS WP1 (games)

• A natural framework in which to study such
  multi-objective optimization problems with local
  payoffs in a non-cooperative network is
  (non-cooperative) game theory. An appropriate,
  game-theoretic concept for the solution is Nash
  equilibrium.

10
FLAGS WP1 (games)

• Roughly speaking, the operating points of a
  non-cooperative network are the Nash equilibria
  of the underlying game these are points where
  unilateral deviation does not help any user to
  improve its performance. Game-theoretic models,
  concepts and techniques will be employed in the
  context of various networking problems such as
  flow control, routing, bandwidth allocation, Web
  access, multicasting and congestion control.

11
FLAGS WP2 (stability)

• We will investigate dynamic situations where
  computing agents move around in a global
  environment trying to communicate and compute
  quickly having only partial knowledge of the
  system. Bad transient behaviour of such global
  systems may be represented by malicious
  adversaries. To study performance under these
  conditions, we will investigate stability and
  fault-tolerance in a novel way.

12
FLAGS WP2 (stability)

• Adversaries can maliciously do the following
• add/inject tasks, with an injection rate on
  non-overlapping paths,
• change the communication graph, by adding or
  removing edges with certain probability.
• Such malicious agents could be very hard to trace
  (and may lead to stability questions that are
  undecidable!).

13
FLAGS WP3 (ad-hoc networks)

• - Todays state-of-the-art techniques are not
  satisfactory (i.e. not efficient, restrictive and
  empirical). In contrast to such static approaches
  we will provide a solid scientific foundation for
  ad-hoc mobile networks, by envisioning networks
  with highly dynamic movement of users and taking
  advantage of the mobile hosts natural movement to
  exchange information whenever mobile hosts meet
  incidentally.
• - Random Walks are used to study correctness
  /performance

14
FLAGS WP3 (ad-hoc networks)

• Smart Dust is a set of a vast number of
  ultra-small fully autonomous devices, with very
  restricted capabilities, that co-operate to
  accomplish a large sensing task.
• We open the research on smart dust from a basic
  algorithmic point of view
• - We provide simple but realistic models for
  smart dust
• - We present smart dust protocols for local
  detection and propagation and perform an average
  case analysis of their efficiency and energy
  consumption.

15
FLAGS A multidiscriplinary approach
The project builds on expertise covering many
aspects of theoretical computer science,
including distributed and parallel computing,
networking and communications, online
decision-making under uncertainty, approximation
algorithms and complexity theory, probabilistic
techniques and combinatorial mathematics.
Furthermore, the new issues arising in the study
of such systems necessitate combining these
techniques appropriately with methods from other
scientific disciplines, such as game theory and
economics, physics and statistics.
16
FLAGS Expected Results

• A unifying scientific framework for
  co-ordination, stability and fault-tolerance,
  motion and communication in complex and dynamic
  global systems.
• A coherent set of design rules and technical
  recommendations for the robust and efficient
  implementation of global systems.
• A set of algorithmic engineering experiments,
  emphasizing on "hard" instances and appropriate
  gross measures.

17
CRESCCO Partners

• University of Patras (GR), coordinator
• Leader Prof. Christos Kaklamanis
• Computer Technology Institute (GR)
• Leader Prof. Paul Spirakis
• University of Geneva (CH)
• Leader Prof. Jose Rolim
• Centre National de la Recherche Scientific,
  Laboratoire I3S (FR)
• Universite de Nice-Sophia Antipolis, Laboratoire
  I3S (FR)
• Leader Prof. Jean-Claude Bermond
• Universitaet zu Kiel (D)
• Leader Prof. Klaus Jansen
• University of Salerno (I)
• Leader Prof. Giuseppe Persiano
• University of Rome Tor Vergata (I)
• Leader Prof. Andrea Clementi

18
CRESCCO Objectives
The integration in modern Information and
Communication Technologies of i) heterogeneous
communication infrastructures (optical, ATM
networks) ii) mobile users accessing these
backbone networks and the Internet, and iii)
dynamic selfish agent entities, results in highly
dynamic, complex, global systems. To design and
implement high-speed, cost-effective, and
reliable communication and computing solutions
for such environments, the project investigates
the bottlenecks and critical issues involved at
the fundamental algorithmic level. In particular,
we focus on the efficient management of scarce
and critical resources such as frequency spectrum
and energy in wireless networks, bandwidth in
optical and ATM networks, as well as CPU time,
space and communication time in dynamic
environments of selfish agents.
19
CRESCCO Work Description

• The research focuses on the following aspects of
  sharing critical resources in global systems
• - The efficient assignment of frequencies and
  call admission control in wireless cellular
  networks - The minimisation of energy
  consumption in wireless networks
• Sharing common resources (like CPU time, space,
  and communication time) on the Internet among
  selfish agent entities - The efficient
  scheduling of bandwidth requests in ATM networks
• - The efficient access to optical bandwidth in
  WDM networks.

20
CRESCCO WP1 (frequencies-energy)

• frequency assignment problems
• complexity results
• efficient approximation algorithms
• energy consumption
• minimum range assignment problems
• to minimize the overall power of the radio
  network

21
CRESCCO WP2 (selfish agents)

• mechanism design
• the impact of coalitions of agents
• the notion of coalitions in equilibrium

22
CRESCCO WP3 (scheduling in networks)

• approximation schemes for
• -scheduling with malleable parallel tasks
• -scheduling ATM requests
• simple and efficient on-line heuristics
• data retrieval in parallel data servers
• lower bounds

23
CRESCCO WP 4 (optical bandwidth)

• Routing, Packing in WDM networks
• NP-hardness results
• integer programming techniques
• efficient approximation algorithms
• - greedy (BFS) techniques for wavelength routing

24
CRESCCO- Expected Results

• A set of algorithmic solutions for the efficient
  management of scarce resources in modern
  communication/computing environments.
• A set of experiments for validating these
  solutions along with worst, random, and real life
  test sets, heuristics, and benchmarks
• Interactions with relevant industry

25
DBGLOBE Partners
26
DBGLOBE- Objectives
Global computing can be seen as a database
problem how to design, build and analyse systems
that manage large amount of data. However, the
traditional database approach of storing data in
monolithic database management systems becomes
obsolete in such environments. In current
database research, data are relatively
homogeneous, exhibit a small degree of
distribution (just a few network sites) are
passive in that they remain unchanged unless
explicitly updated. All these assumptions do not
hold in the global computing world. This creates
the need for new theoretical foundations in all
aspects of data management modelling, storage
and querying. DBGlobe aims at broadening
database management research focus to attack the
issues of mobility, autonomy, incomplete
information, scale, and adaptability that arise
in dynamic environments.
27
DBGLOBE Description of Work (1)
DBGlobe adapts a data-centric approach by
considering each mobile entity as a primary data
store and a mini-server that protects and
encapsulates access to its data. Besides these
"walking" databases of mobile entities, in
DBGlobe, meta-information and services related to
them are maintained in dedicated data stores,
called InfoStations, dispersed throughout the
stationary network. In particular, within the
DBGlobe project, we will develop novel data
management mechanisms along the following key
topics 1. System Architectures There is no
centralized database server, instead, each mobile
object constitutes a database of each own.
28
DBGLOBE-Description of Work (continued)
2. Co-ordination/Data Delivery The objective is
to derive adaptive data delivery mechanisms. 3.
Querying Querying is performed on a multitude of
databases New forms of query languages and
query management paradigms (including filtering,
context-awareness) need to be developed. 4.
Simulation and Proof-of-Concept Prototype to
verify the model, by building a simulator and
implementing a proof-of-concept prototype.
29
DBGLOBE-WP1 (System Architecture)

• To derive appropriate architectures for ad-hoc
  databases of mobile entities. Such architectures
  should be metadata driven. In particular,
• to define what is the appropriate metadata
  information to describe mobile entities
• to derive an appropriate metadata definition and
  manipulation language,
• - to design distribution and replication
  protocols for the DataHolders (that hold
  metadata) and the DataHandlers (that are
  processing entities),
• - to make the architectures dynamically
  configurable and extensible, and
• - to achieve fault-tolerance and availability.

30
DBGLOBE-WP2 (Simulation Environment)

• To design and implement a simulator of dynamic
  environments of co-operative mobile entities. To
  extend the simulator to model the protocols
  advanced through the project. Some of the novel
  issues to be addressed include
• - Modelling the distribution, mobility and data
  of the mobile entities
• - Expressing the interaction among the entities
• - Modelling the ad-hoc creation of databases,
  co-ordination and data acquisition

31
DBGLOBE-WP3 (Data Delivery and Co-ordination)

• To derive adaptive data delivery mechanisms that
  will combine
• - push (transmission of data without an explicit
  request) and pull,
• - periodic and aperiodic , as well as
• - multicast and unicast delivery.
• - To use workflow models and co-ordination
  techniques from multi-agent research to to
  capture the co-ordination among the mobile
  entities and advanced transaction models to
  reason about the correctness of the interaction
  among the entities.

32
DBGLOBE-WP4 (Information Discovery and Querying)

• To derive models and protocols for querying and
  knowledge discovery in dynamic environments of
  co-operative mobile entities. In particular
• Re-define query management to amalgamate
• - Knowledge acquisition,
• - Filtering,
• - Context-awareness,
• - Implicit and continuous evaluation.
• - Develop adaptive query optimisers and execution
  engines to cope with highly unpredictable and
  changeable environments.

33
DBGLOBE-WP5 (Proof-of-concept Prototype)

• To demonstrate through a concrete example the
  idea of ad-hoc databases and the various forms of
  querying
• To provide a concrete example of the workflow
  interaction model and the different data delivery
  mechanisms
• To capture, design and implement a location-aware
  system

34
DBGLOBE Expected results
- Research Results. These results will make
feasible the development of highly dispersed and
massively networked software systems. -
Simulator. The simulator will enable experiments
and performance results for future applications.
It can be used by Research units,
telecommunication companies, software development
groups, for measuring the feasibility and the
performance of methods involving moving objects
and highly distributed data. - Proof-of-concept
prototype. This prototype will be used as a
roadmap of how to use the research results of
DBGlobe in future emerging information systems.
35
SOCS - Partners

36
SOCS Objectives
The aim of this project is 1) to investigate
computational and logical models for describing,
analysing, and verifying individual and
aggregates of computational entities -
(computees) interacting in the context of global,
open and dynamic environments. 2) It further
validates the framework by a series of grounded
controlled experiments using a prototype
demonstrator embodying the formal model. 3)The
results of the project will also provide a
practical basis for the design of classes of
systems and applications which require aggregate
behaviour of computational entities.
37
SOCS Description of Work
The project is divided into three phases formal
models computational models verification and
experiments. 1) During the first phase, the
project undertakes studies that integrate
hypothetical, temporal and argumentation-based
reasoning in order to model logically individual
computees. Then develops a logical framework for
interactions amongst computees. This will
establish interactions amongst computees via
direct communication, whether based on standard
protocols or emerging from individual
communicative behaviours.
38
2) In the second phase of the project,
computational models for the logical models of
individual computees and their interactions will
be developed. 3) To support these computational
models, a complete experimental demonstrator will
be developed which will be tested on scenarios
and examples on varying scales. 4) In the third
phase, the project identifies significant and
desirable properties of computees and their
societies, and proves formally under what
circumstances these properties hold. Results will
be validated using a prototype demonstrator
embodying the formal models.
39
SOCS WP1 (A logic-based model for computees)

• At the level of a single computee we aim at
  modelling
• the knowledge, goals and intentions of a
  computee
• the reasoning abilities of a computee
• the local behaviour that a computee is prepared
  to exhibit within a society
• the global behaviour that a computee may expect
  from other computees within a society.

40
SOCS WP2 (Modelling interactions between
computees)

• WP2 will address the following issues
• identification of models for expressing
  interaction among computees
• formal definition of interaction
• integration of the interaction model with the
  logic-based model for computees (developed in
  WP1)
• capability of adapting interaction to changes.

41
SOCS WP3 (A computational model for societies)

• The principal aim of WP3 is to provide
  computational counterparts to the formal models
  developed by WP1 and WP2. These will pave the way
  to realisations that can be proven correct with
  respect to the formal models. The model we aim at
  will bridge the gap between formal models and
  concrete realisations of societies of computees.
  We will develop
• computational models of individual computees, and
  
• computational models of their integration, via
  interaction, within the societies we envisage.

42
SOCS WP4 (Prototype demonstrator)

• The prototype demonstrator aims at providing a
  testbed for animating what was described
  logically in WP1, WP2, WP3, in order to support
• different reasoning capabilities of a computee
• knowledge, goals, plans and resources of a
  computee
• communication capabilities of a computee
• norms that the computee uses to interact with
  other computees in a society
• adaptive behaviour of a computee
• properties of individual and societies of
  computees.

43
SOCS WP5 (Verifiable properties societies)

• We will explore questions such as
• Will the computees inhabiting a society
  consisting solely of altruistic computees be
  more effective in achieving their objectives than
  those inhabiting a society consisting solely of
  self-interested computees?
• Under what circumstances is negotiation amongst
  the computees a) guaranteed b) not guaranteed to
  terminate?
• Under what circumstances is negotiation
  guaranteed to result in acceptances of offers and
  exchanges of tasks/resources/knowledge in such a
  way that would allow all the computees in the
  society to achieve their objectives?
• What will be the best policies a computee can
  adopt in order to maximise its gains during
  negotiation?

44
SOCS WP6 (Experimentation )

• The aim of WP6 is to test some of the desirable
  and undesirable properties of both individual and
  societies of computees through a series of
  controlled experiments using the prototype
  demonstrator developed by WP4. More specifically,
  we will try out the properties identified and
  verified in WP5 to make predictions that can be
  tested by experimenting with existing scenaria
  identified in WP1 and WP2. The experiments will
  also be carried out with new scenaria whose aim
  is to test and possibly identify unforeseen
  behaviour.

45
Cluster Structure
Models (SOCS)