Contents

1.0 Introduction
Concurrent programming is difficult and the effort is rarely rewarded.

• 1.1 Memory Transactions

2.0 Concurrent Programming Model
Atomic sections do not support interaction with the network or operating system.

• 2.1 Concurrent IO
• 2.2 Parallelism
• 2.3 Compatability
  

3.0 Maintaining Speculative State
Buffering speculative state increases the memory bandwidth requirement of the application.

• 3.1 Speculative State
• 3.2 Immutable Data Structures
• 3.3 Path copying
• 3.4 Binary Trees
• 3.5 Abstract Data Types
• 3.6 Balancing

4.0 Accessing Shared State
A concurrent programming model should have intuitive programming semantics.

• 4.1 Linearizable Objects
• 4.2 Persistent Data Structures
• 4.3 Entanglement
• 4.4 Minimum Spanning Tree

5.0 Concurrency Control
A scalable concurrent system should implement distributed concurrency control.

• 5.1 Distributed Concurrency Control
• 5.2 Serialisability
• 5.3 Confluence

6.0 Contention Management
Strong progress guarantees eliminate the need for contention management.

• 6.1 Progress and Contention Management
• 6.2 Non-blocking Algorithms
• 6.3 Producer Consumer Queue
• 6.4 Distribution and Scheduling
  

7.0 Conclusion
Immutable data and distributed concurrency control are co-dependent.

• 6.1 Conclusion and Future work