PRINCIPLES OF

1
PRINCIPLES OF DISTRIBUTED COMPUTING
Instructor Stefan Dobrev E-mail
sdobrev_at_site.uottawa.ca Office SITE 5043 Office
hours Tuesday 1500-1700 Course page
http//www.site.uottawa.ca/sdobrev/CSI5308
2
Course objective

• give you some background in distributed
  computing
• get familiar with basic problems studied in
  distributed computing
• learn basic distributed algorithms and
  algorithmic techniques
• learn how to evaluate distributed algorithms
• provide deeper insight into selected areas of
  distributed computing

3
Textbook and Reading

• there is no single course textbook
• selected sources
• Nicola Santoro, Design and Analysis of
  Distributed Algorithms
• http//www.scs.carleton.ca/santoro/DADA/Not
  es.html
• Gerard Tel Introduction to Distributed
  Algorithms, 2nd edition, Cambridge University
  Press, 2000, ISBN 0521794838
• Nancy Lynch Distributed Algorithms, Morgan
  Kaufmann, 1996, ISBN 1-55860-348-4

4
Topics Covered

• Distributed Environment
• Basic Problems and Simple Solutions
• Computation in Trees
• Leader Election and MST construction
• Termination Detection
• Global Snapshots
• Synchrony in Networks
• Fault Tolerant Computing
• Routing

5
Topics in Detail I

• 1. Distributed Environment
• formal model, axioms
• behaviour/protocol, events, actions, causality
• complexity measures
• restrictions, structural knowledge
• Why?
• To know what we are talking about.

6
Topics in Detail II

• 2. Basic Problems and Simple Solutions
• broadcasting, wake-up
• leader election, spanning tree construction ,
  DFS
• flood and echo algorithms, counting, naming
• Why?
• These are basic problems which are used as a
  building block in many more complex distributed
  protocols. In many cases the simple and naïve
  solutions are sufficient, or cannot be improved
  anyway.

7
Topics in Detail III

• 3. Computation in Trees
• saturation technique and its applications
• Why?
• Simple and powerful technique. Often a
  computation for the whole graph is reduced to a
  computation on its spanning tree.

8
Topics in Detail IV

• 4. Leader Election and Minimum Spanning Tree
  Construction
• in rings
• in other special topologies (mesh, torus,
  hypercube)
• in arbitrary networks
• discussion on approaches to leader election
• Why?
• Fundamental symmetry breaking technique, involved
  in many places as a building block. Studying it
  on different topologies gives a lot of insight
  into the effects of structural information on
  complexity of distributed algorithms.

9
Topics in Detail V

• Termination Detection
• Global Snapshots
• Synchrony in Networks
• counting technique, communicators, election in
  synchronous rings
• synchronizer algorithms
• Fault Tolerant Computing
• impossibility of asynchronous consensus
• solutions synchrony, randomization, failure
  detectors
• Routing
• distributed maintenance of routing tables
• routing with compact routing tables
• routing in geometric networks

10
Grading (tentative)

• 35 4 assignments (A)
• 30 midterm (M)
• 35 project (P)

11
Assignments (tentative)

• 4 assignments for a total of 35 points
• one assignment compiling course notes
• one assignment theoretical, to prepare you for
  the midterm
• two programming assignments, using a simulator
  of distributed system

12
Midterm (tentative)

• Midterm (30 points)
• about end of October, beginning of November
• all topics covered up to that moment
• open book, 90-120 minutes
• possible questions
• write an algorithm for a variant of a problem
  studied on the lectures
• evaluate complexity of a distributed algorithm
• justify why the provided algorithm does/does not
  work

13
Project (tentative)

• Project (35 points)
• individual or in groups, depending on the number
  of students
• written report 45-60 min in-class presentation
  (end of semester)
• topics will be posted during September
• several students/groups can work on the same
  topic