Key Consistency Models to Know for Parallel and Distributed Computing

Consistency models are crucial in parallel and distributed computing, ensuring data accuracy across multiple processes. They define how operations appear to users, balancing performance and reliability, from strong consistency to eventual consistency, each serving different system needs.

1. Strong Consistency

   • Guarantees that all operations appear to occur instantaneously at some point between their start and end.
   • Ensures that once a write is acknowledged, all subsequent reads will reflect that write.
   • Often requires synchronization mechanisms, which can impact performance in distributed systems.

2. Eventual Consistency

   • Guarantees that if no new updates are made, eventually all accesses will return the last updated value.
   • Allows for temporary inconsistencies, making it suitable for systems that prioritize availability and partition tolerance.
   • Commonly used in distributed databases and systems like DNS and social media feeds.

3. Causal Consistency

   • Ensures that operations that are causally related are seen by all processes in the same order.
   • Non-causal operations may be seen in different orders by different processes.
   • Balances the need for consistency with the flexibility of concurrent operations.

4. Sequential Consistency

   • Requires that the result of execution is the same as if all operations were executed in some sequential order.
   • The sequential order must respect the program order of each individual process.
   • Provides a more relaxed model than strong consistency while still maintaining a global order.

5. Linearizability

   • A stronger form of consistency that ensures operations appear to take effect instantaneously at some point between their start and end.
   • Guarantees that the results of operations are consistent with a single, global timeline.
   • Often used in systems requiring strict consistency, such as distributed databases.

6. Read-your-writes Consistency

   • Guarantees that a process will always see its own writes.
   • Ensures that once a write is made by a process, subsequent reads by that process will reflect that write.
   • Useful in user-facing applications where users expect to see their own updates immediately.

7. Session Consistency

   • A form of consistency that guarantees that within a session, a user will see a consistent view of the data.
   • Allows for different users to see different views of the data, as long as each user’s view is consistent during their session.
   • Balances user experience with system performance in distributed environments.

8. Monotonic Read Consistency

   • Ensures that if a process reads a value, any subsequent reads will return that value or a more recent one.
   • Prevents processes from seeing "stale" data after they have seen a more recent value.
   • Important for applications where users expect to see progressively updated information.

9. Monotonic Write Consistency

   • Guarantees that writes by a single process are seen in the order they were issued.
   • Prevents scenarios where a later write is seen before an earlier one by the same process.
   • Helps maintain a logical flow of updates in applications that require ordered operations.

10. PRAM (Pipelined Random Access Memory) Consistency

    • A model that allows for concurrent reads and writes while ensuring that operations appear to be executed in a pipelined manner.
    • Provides a balance between performance and consistency in parallel computing environments.
    • Useful in systems where multiple processors need to access shared memory efficiently while maintaining some level of consistency.