CSR: Medium: Collaborative Research: Workload-Aware Storage Architectures for Optimal Performance and Energy Efficiency

点击翻译按钮获取中文摘要

The most significant performance and energy bottlenecks in a computer areoften caused by the storage system, because the gap between storage deviceand CPU speeds is greater than in any other part of the machine. Big dataand new storage media only make things worse, because today's systems arestill optimized for legacy workloads and hard disks. The team at StonyBrook University, Harvard University, and Harvey Mudd College has shown thatlarge systems are poorly optimized, resulting in waste that increasescomputing costs, slows scientific progress, and jeopardizes the nation'senergy independence.First, the team is examining modern workloads running on a variety ofplatforms, including individual computers, large compute farms, and anext-generation infrastructure, such as Stony Brook's Reality Deck, amassive gigapixel visualization facility. These workloads produce combinedperformance and energy traces that are being released to the community.Second, the team is applying techniques such as statistical featureextraction, Hidden Markov Modeling, data-mining, and conditional likelihoodmaximization to analyze these data sets and traces. The Reality Deck isused to visualize the resulting multi-dimensional performance/energy datasets. The team's analyses reveal fundamental phenomena and principles thatinform future designs.Third, the findings from the first two efforts are being combined to developnew storage architectures that best balance performance and energy underdifferent workloads when used with modern devices, such as solid-statedrives (SSDs), phase-change memories, etc. The designs leverage the team'swork on storage-optimized algorithms, multi-tier storage, and new optimizeddata structures.