Distributed Systems 2

1
Distributed Systems 2

• Distributed Processing

2
Process

• A process is a logical representation of a
  physical processor that executes program code and
  has associated state and data.
• Can have separate address space
• Processing created by running app or language
  construct (e.g.fork)
• Operating system carries out a context switch
• Processes that share data lightweight processes
  or threads

3
Co-operation

• Why is cooperation required ?

4
Synchronization

• On the same computer fairly easy. Shared clock
  and can share memory.
• Semaphores are one method
• WAIT(sem-name)
• SIGNAL(sem-name)
• Semaphores can be used in the solution of class
  distributed processing problem the Dining
  Philosophers

5
Dangers of Synchronization

• Starvation
• Deadlock

6
Other kinds of interprocess synchronization

• Eventcounts
• ADVANCE(count-name)
• READ(count-name)
• AWAIT(count-name, value)
• Similar to the coordination at a deli counter

7
Inter Process Communication

• IPC for short
• One example is the synchronous filter mechanism
  think of the unix pipe
• One process writes to the pipe until it is full
• Unidirectional and bound to source and target
• Not secure
• Can also used a Named Pipe

8
Files

• Files can be used for process communication
• Can handle large volumes of data
• What kind of problems do you think you might run
  into to ?

9
Shared Memory

• Local processes can use shared memory area.
• No inherent means of synchronization so
  semaphores may be used
• Not suitable for large volumes of data
• Think of a clipboard

10
Message Queues

• Process can read and write to a named queue
• Synchronization is built in
• Only suitable for small amounts of data

11
Structuring a distributed system

• Master-slave
• Client/server
• Peer-to-peer
• Group
• Distributed object
• Multimedia stream

12
Samples

• http//www.cs.cf.ac.uk/Dave/C/