徐志偉 李曉維 吳承勇 黃鎧 宮云戰
摘 要:該報告研究新興戰略性應用(如物聯網、三網融合、云計算等)的后端高通量處理的新原理、新方法和核心技術。主要研究內容包括由控制或數據驅動轉向請求驅動的新原理、大規模可擴展微處理芯片的新結構及其相關核心技術,高通量低能耗的新型計算機體系結構、支撐其高效能執行的新型運行時管理系統和新的軟件支持環境,虛擬化資源能力的部署、調度、管理和優化技術,為按需聚合、智能協同的高通量應用提供可保障的系統能力;同時研究各個階段的可靠性設計和驗證的新機制,包括硬件級的芯片和系統的可靠性機制,軟件級的容錯、缺陷調試和程序分析驗證等,以及高通量億級并行系統的測試程序、健壯性、性能自動推演和性能瓶頸自動發現等技術。通過該研究,提出并實現有助于解決重大科學問題的高通量計算系統的創新結構和設計方案。
關鍵詞:高通量 計算系統 構建 云服務 應用
Abstract:The aim of this research is to advance hyper-parallel processing (HPP) and highly energy-efficient computer architecture and design methodology, and apply these methods into cloud computing applications. Our research topics focus on:(1)how to design novel HPP computer architecture, parallel algorithm, compiler and operating system.(2)how to design highly energy-efficient architecture and hardware that achieve reasonable performance at the cost of low power consumption.(3)how to design reliable and performance predictable systems based on massive error-prone components.(4)how to apply these proposed method into cloud computing effectiviely. We hope with this rpoject, we can propose two navel systems, which are many-core microprosessor and computing sysytem with high energy-efficiency, high reliability, and scalablity. Meanwhile, our project will help improve buliding advanced data center and foster national economy on informantion development.
Key Words:Proposal;High-Thoughput;973
閱讀全文鏈接(需實名注冊):http://www.nstrs.cn/xiangxiBG.aspx?id=42924&flag=1