Towards Next Generation Panel at SAINT 2002

1
Towards Next GenerationPanel at SAINT 2002

• Gul Agha
• Professor of Computer Science and Computational
  Science and Engineering
• University of Illinois at Urbana-Champaign
• http//www.osl.cs.uiuc.edu

2
Research Directions

• What are the programming models for Applications
  on the Internet?
• Massively parallel, distributed and mobile.
• Multi-agent Systems
• Actors
• Network Embedded Systems
• Integration of continuous and discrete processes.

3
Large Scale Actor Systems

• Reasoning about large-scale agent systems.
• Asynchrony
• Autonomous agents that communicate
  asynchronously.
• Modularity
• Concurrent components each with several agents.
• Locality and Non-interference
• Message are the only means of information flow.
• Mobility
• Reconfiguration and extension of systems.

4
Formal Methods

• Probabilistic reasoning about distributed and
  mobile real-time systems
• Techniques for analysis of algorithms.
• Performance models.
• Models must account for
• Stochastic duration of actions.
• Probabilistic behavior of node or system.
• Uncertainty in the environment.

5
Metrics for Large-Scale Systems

• Probabilities on evolution of system.
• Quantifying robustness, stability, timeliness,
• Summations over execution paths for statistical
  metrics..

6
Models of Time

• Global synchronous wall clock
• Synchronization is too tight
• Too detailed an execution model
• Asynchronous, distributed time
• Vector clocks are too expensive
• Application behavior is complicated
• Need a more expressive model of time
• Notion of distance and distribution.
• Space-Time cone of causal influence.

7
Causality and Time

t
Future causality
R
P
Q
Not Causally related
x
S
Past causally connected
y
8
Agents with Bounded Resources

• Each agent has bounded resources.
• Dynamic binding of agents to computational and
  network resources.
• Control of Multi-Agent Systems by bounding
  resources
• Stability
• Efficiency
• Optimizing search
• Preventing denial of service attacks

9
Coordination Algorithms

• Large-scale systems where
• Nodes and links have variable capabilities.
• Real-time requirements in the absence of a
  predefined global clock.
• Faults are common.

10
Example Global Snapshot

• Consistent with causal relations.
• Real-time constraints
• Duration of snapshot skew.
• Completion time of snapshot.
• Scalable
• Minimize number of messages.
• Dynamic topology of nodes and channels.
• Notion of approximation in snapshots
• Farther apart nodes have larger skews.

11
New Directions

• Hybrid Systems
• Network Embedded Systems
• More complex middleware
• Naming
• Computational Reflection
• World-Wide Computer
• Mobility
• Grid-based computing