Transparent serverless execution of Python multiprocessing applications

Access transparency means that both local and remote resources are accessed using identical operations. With transparency, unmodified single-machine applications could run over disaggregated compute, storage, memory resources. Hiding the complexity of distributed systems through would have great benefits, like scaling-out local-parallel scientific flexible in Cloud. This paper presents a performance evaluation where we assess feasibility access state-of-the-art Cloud for Python multiprocessing applications. We interfaced module with an implementation transparently runs processes on serverless functions uses in-memory data store shared state. To evaluate four applications: Uber Research's Evolution Strategies, Baselines-AI's Proximal Policy Optimization, Pandaral.lel's dataframe, ScikitLearn's Hyperparameter tuning. compare execution time scalability same application running our library, libraries large VM. For equal resources, efficiently message-passing abstractions achieve comparable results despite significant overheads communication. Other shared-memory intensive do not perform due to high latency. The show Python's library design is enabler towards transparency: legacy efficient can scale beyond VM limited increased parallelism without changing underlying code or architecture.