Lecture 7

• Lecture 7
  • Video
  • Slides
  • Need for a Model
  • Message Passing model for IPC
  • Axioms for reliable channel
  • Channel

Video

link

Slides

link

Need for a Model

• models are abstractions

Message Passing model for IPC

• Representation of Distributed System
  • represented as a graph
  • node is a sequential process
  • edge is communication channel
  • graph is directed

• internal
  • proces performs action in its own address space which results in modificn of local variables
  • so no affect on other processes
• communication action
  • when process sends msg to another process or rcv from someone else
• input
  • reads data from sources external to the system
  • eg: sensors hai, and processes are interacting with it
• output
  • again external to system
  • say process is monitoring boiler temperature and raises an alarm when needed
  • part of system which is external = universe

Axioms for reliable channel

• axiom = a stmt of propisition which is self-evidently true

• why violation of axiom 1 at data link or transport layer
  • could be due to congestion in n/w
  • retransmission hai transport layer pe to correct it
• absence of predifened upper bound
  • it weakens computation model
  • sometimes we might be able to specify an upper bound
  • eg is slot time
• why not satisfied by datagram service
  • bcz packets can be out of order

• 2 possible ways to abstract channel capacity
  • infinite capacity
    • sender sends as much as it wants
    • and channel won’t ever block
    • it is irrespective of speed of processing of rcvr
  • finite
    • blocking is possible
    • so dropping could be there

Channel

• is kind of sharing data object
• but not like memory
• u cannot write to a location, only append
• cannot un-send
• cannpot read any element, only headers
• no functionality to check if channel is empty or not