Accelerating Scientific Workflows with Tiered Data Management System
Peng Cheng, Yutong Lu, Yunfei Du, Zhiguang Chen · 2018
The explosion of scientific data generated from large-scale simulations and advanced sensors makes scientific workflows more complex and more data-intensive. Supporting these data-intensive workflows on HPC systems presents new challenges in data management due to their scales, coordination behaviors and overall complexities. In this paper, we present Tiered Data Management System (TDMS) to accelerate scientific workflows on tiered storage architecture. TDMS utilize the throughput and capacity characteristics of each storage tier for efficient data sharing. Moreover, TDMS provides customized data management strategies for different workflow data access patterns to make full use of advantages of different storage tiers. We build a prototype and deploy it on representative HPC system. We evaluate the performance of TDMS with realistic workflows and the experiments show that the customized data management strategies can optimize the I/O performance and provide 1.6x speedup for data-intensive workflows compared with Lustre file system.