Distributed systems Spring 2007

LH 5  Tanenbaum Ch6

1. a) In most implementations of the consistency model "(Eager) Release" the
      shared variables are synchronized on a release but not on an acquire.
      Why is acquire needed at all then?

   b) The consistency models that use a synchronization variable have
      specific consistency rules for the use of these synchronization
      variables. What are these rules in each case? Would any alternative
      specifications have been possible?

   c) What are the essential differences between the models "weak consist-
      ency", "release consistency" and "entry consistency"?

2. A working group belonging to a large global project has to travel
   quite a lot all around the world. All the workers have portable
   computers, and they also have GPRS-level access to the project's file
   and database servers (the most important of these are replicated in
   different parts of the world).

   In order to improve its internal cooperation the group wants to have
   a "shared desk" with
  - a bulletin board and
  - a repository for active files (reports etc, updatable by everybody).

  Which consistency model should be applied in this application? Go
  through all consistency models (data centric and client centric) and
  for each evaluate the benefits and costs involved (implementation and
  maintenance, the (in)conveniences for the users etc; however, there is
  no need think in terms of money).

3. To continue with the exercise 2 above: 
   How could the model(s) you propose be implemented?


4. Estimate the benefits and drawbacks of epidemic communication in
   - wireless, mobile ad hoc network
   - implementation of routing in ad hoc networks
   - implementing routing in the internet.
  What kind of push/pull solution would be rational?
  Reference: MANET
 

The last 2 home exercises will be published during the weekend.