Exploring Memory Hierarchy to Improve Scientific Data Read Performance

Abstract

Improving read performance is one of the major challenges with speeding up scientific data analytic applications. Utilizing the memory hierarchy is one major line of researches to address the read performance bottleneck. Related methods usually combine solide-state-drives(SSDs) with dynamic random-access memory(DRAM) and/or parallel file system(PFS) to mitigate the speed and space gap between DRAM and PFS. However, these methods are unable to handle key performance issues plaguing SSDs, namely read contention that may cause up to 50% performance reduction. In this paper, we propose a framework that exploits the memory hierarchy resource to address the read contention issues involved with SSDs. The framework employs a general purpose online read algorithm that able to detect and utilize memory hierarchy resource to relieve the problem. To maintain a near optimal operating environment for SSDs, the framework is able to orchastrate data chunks across different memory layers to facilitate the read algorithm. Compared to existing tools, our framework achieves up to 50% read performance improvement when tested on datasets from real-world scientific simulations.

Toward Scalable and Asynchronous Object-Centric Data Management for HPC

“…The data in the burst buffer are then flushed to HDDs when the size is more than half of the burst buffer capacity. [31], [32] proposes a framework that is able to utilize the deep memory hierarchy resources to improve read performance as well as runtime data sharing. In our previous work, Data Elevator [33] provides automatic data movement across multi-levels of deep storage systems.…”

ARCHIE: Data Analysis Acceleration with Array Caching in Hierarchical Storage

“…A few recently proposed cache management systems have the prefetching feature. Eley [33] and [34] prefetches the input data of applications into SSD to reduce the latency of reading data inputs. Byna [2] and Tang [24], [23] propose prefetching algorithms that fetch data based on the historical access patterns of flat files.…”