Distributed Systems, Spring 2007

Home exercise 3

The essay should be returned by e-mail to 
Lea.Kutvonen@cs.helsinki.fi, mika.karlstedt@cs.helsinki.fi 
as an attached file (preferably in pdf format, with the name 
[last name].[format type]) at the latest on Thursday 22.2. at 23.59.
       
1. Coda checks the validity of file replicas using a version vector CVV,
   which contains the number of updates in the replicas. Let us assume
   that the replicas of the file F are stored on the servers A and B.
   The connection between these servers is often down, even for longer
   periods. The user has a portable workstation, and the AVSG for it is
   sometimes only the A server, sometimes only the B server.
   a) The workstation wants to upload the updated file F to the servers.
      Describe the updating of the CVV vectors in the case where both
      servers are available.
   b) Let us assume that in the server A the CVV of F is (3,2) and in the
      server B it is (2,4). How many updates of F have been made? How many
      of them are in both replicas? How many are only in A, how many only
      in B?
   c) Why is CVV=(2,3) not possible in A? If CVV=(3,2)in A, what are the
      possible values of CVV which can be found in B? How does the work-
      station know which of the replicas contains the most current data?
      How does the system update its replicas to contain only up-to-date
      data?
   d) Must the F replica in the workstation contain a CVV?
   e) What is the relationship between CVV and the timestamp used to
      implement causal ordering? Could CVV be used for that purpose?