It sis te mas sisgesa

Fundamental Distributed System Models

Adapted from From DistributedSystems: Concepts and Design, Coulouris, Dollimore and Kindberg Edition 4, © Addison-Wesley 2005

Fundamental Models
Fundamental Models – Description of properties that are present in all distributed architectures.
Interaction Models – Issues dealing with the interaction of process such as performance and timing of events. Failure Models – Specification of faults that can be exhibited by processes and communication channels. Security Models – Threats to processes and communication channels

1

Interaction Model Challenges
Distributed systems are composed of cooperating processes:
Generally unpredictable timing and rate of message transmission delivery between processes. Processes maintain their own private state There is no single global notion of time.
Clock drift rate – relative difference process clocks differ from a perfect reference clock.

Performance Considerations
Latency – delay between the transmission start of a message and its receipt.
Time taken by the actual message transmission will vary with load of message traffic and time required by OS services to process messages.

Bandwidth – amount of information that can be transmitted over a computer network in a given amount of time.

2

Two Interaction Model Flavours
Synchronous Model
Boundaries known for time to execute step, message transmission and clock drift rate. Timeouts typically used to detect failures Ineffective resource sharing

Asynchronous Model
No assumptions made on process time, message transmission or clock drift. Event ordering cannot be dependent on time More opportunity for resource sharing, but much more complex design.

Real-time ordering of events send X 1 m1 2 receive send 3 m2 receive Physical time receive 4 receive Y

send Z receive receive m3 t1 t2 m1 m2

A

receive receive receive t3

• Delivery time of messages cannot be predicted since clocks cannot be perfectly synchronized across a distributed system •