Distributed Computing and Systems

1
Distributed Computing and Systems

• What does this mean ?
• Applications running on several computers
  distributed within the same room of over long
  distance
• Systems consisting of a large collection of
  computing resources that provide a simplified
  service interface which hides the distribution of
  the resources
• Middleware software that simplifies the creation
  of distributed applications or manages the
  distributed system of resources

2
Distributed Computing and Systems

• Particular functions provided by distributed
  computing systems
• Reliability and Fault tolerance automatic
  recovery from failure of an individual resource
• Load sharing e.g. Googles service runs on a
  distributed system

3
Distributed Computing and Systems

• Difficulties of designing distributed systems
• Need for well-defined communication protocols (at
  the different levels of the protocol hierarchy)
  between the different system components
• Complexity of understanding all the possible
  execution sequences, due to the concurrency
  involved

4
Distributed Computing and Systems

• Application areas
• Peer-to-peer systems
• Communication systems (IP-telephony,
  teleconferencing, collaborative work)
• E-commerce
• Distributed data bases
• Management of distributed systems (networks,
  computing resources, automated factories, etc)

5
Distributed Computing and Systems

• Professors involved
• Bochmann, Boukerche, Dobrev, Flocchini, Groza,
  Ionescu, Jourdan, Karmouch, Miri, Nayak,
  Stojmenovic, Ural

6
Distributed Algorithms
Flocchini
Classical distributed computations

• The importance of local information
• Impact on communication complexity of
  information locally available at the nodes (e.g.,
  informative port-numbers).
• - applications of graph algorithms
• Computing in presence of faults
• Study of distributed algorithms tolerant to
  various failures (e.g. Point of failure
  re-routing, majority voting systems, )

7
Distributed Algorithms
Dobrev and Flocchini
Algorithms for Mobile Agents

• Network decontamination
• An intruder moves contaminating nodes. The
  goal deploy a (possibly small) team of mobile
  agents to decontaminate the network.
• Applications in security (virus location).
• Computing in a dangerous network
• Nodes are affected by permanent, static,
  malicious faults.
• An agent passing by a dangerous node is not
  allowed to exit.
• The goals
• 1) deploy a team to locate the faults (without
  loosing too many agents)
• 2) perform correct computations in spite of the
  faults.

Study of various algorithmic issues that arise
in these contexts
8
Distributed Algorithms
Jourdan and Ural
Modeling and testing

• Objective Infer models of distributed systems on
  which to efficiently perform complex tests
• Main research questions
• Loop inference from a set of observations of a
  distributed systems, infer the design and the
  possible repetitive sub functions
• Distributed testing generating test sequences
  including the necessary coordination messages
  between testers to avoid controllability and
  observability problems (e.g. reduce the need for
  coordination messages, reduce the number of
  coordination channels, avoid them altogether
  etc.)
• Input/Output Partial Orders automata specify a
  distributed system with a partial order automata
  and generate test sequences for it

9
Distributed Algorithms
Ionescu
Some ongoing projects

• Patterns for Autonomic Computing
• with IBM CAS
• Secure Content Routing for B2B SOA
• funded by ORNEC
• Collaborative Web-based Desktops
• funded by NSERC

10
Distributed Algorithms
Boukerche

• High-Performance Computing
• Project1 Parallel strategies for the local
  biological DNA sequence alignment in a cluster of
  workstations
• Distributed Interactive Simulation System
• Project1 Multicast group management in
  large-scale distributed interactive simulation
  systems
• Project2 Resource control for large-scale
  distributed simulation system over loosely
  coupled domains
• Project3 Real-Time distributed simulation
  systems

11
Application area Sensor Networks

• Distributed information collecting system made of
  a large number of tiny, limited capacity nodes
• Self-organized, cooperative, multi-hop
  communication
• Mostly directional data flow towards more capable
  sink nodes
• No mobility
• Desired features Easy to deploy, survivable,
  robust, scalable

12
Sensor Networks Protocol Design Perspective

• Network size
• Large network Scalability problem. Protocols for
  small scale networks cannot be directly used
• Communication pattern
• Sensor networks many-to-one, only sporadic
  transmissions, unique destination
• Mobility
• Sensor nodes are stationary
• Small routing overhead, no or little route
  request broadcast
• Important implications over many network layers
  including security

13
Distributed Algorithms
Miri
Security challenges in Sensor Networks

• Intrusion/anomaly detection
• Secure routing
• Lightweight cryptography

14
Distributed Algorithms
Bochmann
Some ongoing projects

• Peer-to-peer networks optimization and load
  balancing taking geographical location of servers
  into account
• Distributed systems management for high-speed
  networks and Grid applications User-Controlled
  Lightpath Provisioning (collaboration with CRC
  (Ottawa), i2Cat (Barcelona) and Inocybe
  (Montreal) see www.uclp.ca)