Scalable Non-Blocking Atomic Snapshot Algorithm in MPI Remote Memory Access Model

This paper describes a scalable, non-blocking method for performing atomic snapshots in a distributed memory system using MPI Remote Memory Access (RMA). The algorithm achieves atomicity through MPI_Compare_and_swap operations, ensuring correctness under concurrent updates. It distributes workload among MPI ranks and avoids locks or global synchronization, enabling high scalability for large datasets and high numbers of processes.The workload is evenly distributed among processes, minimizing contention and achieving linear scalability with increasing dataset size and number of processes.The algorithm's theoretical foundation guarantees atomicity and linearizability, while its design avoids locks and global synchronization, ensuring efficient execution in high-concurrency environments. Experimental evaluation demonstrates its scalability and robustness, with minimal communication overhead even under heavy workloads. This work provides a practical and efficient solution for distributed systems requiring consistent snapshots, contributing to advancements in high-performance computing and large-scale data analytics.