Introduction to Distributed Systems EEE466

1
Introduction to Distributed SystemsEEE466

• Dr Scott Knight
• Royal Military College of Canada
• Electrical and Computer Engineering
• Knight-s_at_rmc.ca
• 1-613-541-6000 ext. 6493

2
References

• Coulouris, Dollimore and Kindberg, Chapter 2

3
Topics

• Distributed Systems Models
• Interaction Model
• Failure Model
• Security Model

4
Review

• Client-Server Model
• Peer-to-peer
• Mobile code
• Network computers
• Thin-clients
• Note All these models involve multiple
  interacting processes

5
Interaction Model

• The distributed algorithm concept
• Steps to be taken by the programs involved
• Messages to be exchanged
• Conveyance of information
• Co-ordination of activity
• Example of phone system as state machines

6
Interaction Model

• Significant factors affecting interaction of
  distributed processes
• Communications performance
• Cannot maintain a notion of global time

7
Interaction Model

• Performance of Communications Channels
• Latency
• The delay between one process sending a message
  and an other process receiving the message
• Tx time on the link (e.g. land fiber vs.
  sattelite)
• Delay in accessing the network
• Operating system processing delays
• Bandwidth
• Amount of information passed per unit time
• Jitter
• Variation in delay/latency

8
Interaction Model

• Difficulty with global time
• Many individual computer clocks
• Clocks drift at different rates
• Mitigations
• Time synchronization messages
• GPS time (e.g. 1 µsec accuracy)

9
Interaction Model

• Two time models
• Synchronous distributed systems
• Upper and lower bounds on process execution time\
• Upper bound on message transmission time
• Upper bound on clock drift rate
• Asynchronous distributed systems
• No bounds on these three things
• Note Without a global clock perfect event
  ordering is not possible

10
Failure Model

• Processes can fail
• Communications can fail
• Failure Types
• Omission failures
• Arbitrary failures
• Timing failures

11
Failure Model

• Omission Failures
• Process stops
• Crash when we cannot be sure of stoppage
• Fail-stop when we a certain the process has
  stopped (only possible in synchronous systems)
• The impossibility of reaching agreement in the
  presence of failures
• Proof

12
Failure Model

• Arbitrary Failures
• A process may or may not do what is required
• A process may or may not do additional unintended
  things
• Communications may corrupt messages, messages can
  be fabricated, messages can be duplicated
• Worst kind of failures
• If we can, we convert these to omission failures

13
Failure Model

• Timing Failures
• In Synchronous distributed systems these are
  failures of one of the three basic assumptions,
  I.e
• Bounds on processing time
• Bounds on communications delay
• Bounds on clock drift

14
Security Model

• Protecting Objects and the Reference Monitor
  concept
• Principals
• Objects
• Access rights

15
Security Model

• The Enemy can
• Observe
• Alter/stop messages
• Copy and replay messages
• Inject new messages
• DoS (denial of service)

16
Security Model

• Security Mechanisms
• Cryptography
• For link confidentiality
• For link integrity
• Authentication
• Principal identification
• To address man-in-the-middle attacks