Research

From “Small computing” such as mobile and embedded system to “Big computing” such as large-scale data center, modern computer system has evolved to suit a variety of socially demanding computing environments. High Performance Computing (HPC) Lab conducts research on exploring a variety of research into memory-based next-generation computer architecture and systems for rapidly changing computing environments. Recently, in order to solve the performance gap between processor-memory-storage, which is getting worse due to the use of memory/data-centric applications such as big data processing, artificial intelligence, etc., we focus on the new memory-storage systems and intelligent data management techniques using machine learning/reinforcement learning .

Some of the on-going research topics are listed below.

Memory-Storage system based on Machine Learning Workload Analysis

Current big data applications show low spatial/temporal localities and memory-/IO-bound characteristics in conventional memory-storage hierarchy system. In this research, we propose workload-aware memory-storage systems to fit in with performance requirements. We are applying machine learning algorithms to analyze memory-storage request patterns and using it to make memory-storage resource management decisions.

Publications:

  • Pattern Analysis based Data Management Method and Memory-disk Integrated System for High Performance Computing, 2020, FGCS

  • Access pattern-based high-performance main memory system for graph processing on single machines, 2020, FGCS

Intelligent Data Management for Hybrid Memory System

The easiest way to solve issues of data-centric applications is to use a large amount of fast working memory. To design a large-scale main memory with low access latency, we propose a hybrid main memory system using the next generation non-volatile memories (NVMs) such as PCM (phase-change memory) and its management techniques. In this research, we focus on improving performance and energy consumption.

Publications:

  • Q-Selector-Based Prefetching Method for DRAM/NVM Hybrid Main Memory System , 2020, Electronics

  • Regression Prefetcher with Preprocessing for DRAM-PCM Hybrid Main Memory , 2018, IEEE Computer Architecture Letters