tf-Darshan: Understanding Fine-grained I/O Performance in Machine Learning Workloads

Abstract

Machine Learning applications on HPC systems have been gaining popularity in recent years. The upcoming large scale systems will offer tremendous parallelism for training through GPUs. However, another heavy aspect of Machine Learning is I/O, and this can potentially be a performance bottleneck. TensorFlow, one of the most popular Deep-Learning platforms, now offers a new profiler interface and allows instrumentation of TensorFlow operations. However, the current profiler only enables analysis at the TensorFlow platform level and does not provide system-level information. In this paper, we extend TensorFlow Profiler and introduce tf-Darshan, both a profiler and tracer, that performs instrumentation through Darshan. We use the same Darshan shared instrumentation library and implement a runtime attachment without using a system preload. We can extract Darshan profiling data structures during TensorFlow execution to enable analysis through the TensorFlow profiler. We visualize the performance results through TensorBoard, the web-based TensorFlow visualization tool. At the same time, we do not alter Darshan's existing implementation. We illustrate tf-Darshan by performing two case studies on ImageNet image and Malware classification. We show that by guiding optimization using data from tf-Darshan, we increase POSIX I/O bandwidth by up to 19% by selecting data for staging on fast tier storage. We also show that Darshan has the potential of being used as a runtime library for profiling and providing information for future optimization.

Caracterização de Operações de E/S por meio da Análise de Logs

“…O Darshan é uma ferramenta de caracterizac ¸ão de E/S que coleta informac ¸ões de aplicac ¸ões sem afetar consideravelmente o seu desempenho. Desenvolvida no laboratório de Argonne, é muito popular em estudos que utilizam seus logs para analisar cargas de trabalho de supercomputadores e caracterizar o comportamento das operac ¸ões de E/S [Pavan et al 2019, Devarajan et al 2020, Chien et al 2020.…”

Improving I/O Performance via Address Remapping in NVMe Interface

“…As emerging big data and machine learning applications produce and process a large amount of data, the storage performance is becoming more and more important [1], [2], [3]. To improve storage performance, flash-based solid-state drives (SSDs) have been widely adopted in both industry and academia as they provide higher bandwidth and lower latency compared with the existing hard disk drives (HDDs) [4].…”