very high speed computing system 中文意思是什麼
very high speed computing system
解釋
超高速計算系統-
Although gpu has a very high computing speed, algorithms implemented in cpu cannot be put to execute in gpu directly because of the discrepancy in instruction execution manner of the two. gpu ' s architecture is a high parallel simd instruction set system. to reimplement algorithms insufficient to run on cpu with programmable graphics hardware, it has to reconsider the data structures and procedures to implement them to make full use of the
雖然gpu具有非常高的計算速度，但並不能將以前在cpu中實現的演算法直接放到gpu中來執行，這是因為gpu的指令執行方式和cpu不一樣， gpu的體系結構是一種高度并行的單指令多數據（ simd ）指令執行體系，所以要在可編程圖形硬體上實現在cpu中效率不高的演算法，就需要重新制定演算法實現的數據結構和步驟，以充分利用gpu并行處理體系結構帶來的性能優勢。 -
As the application of distributed computing is used much more widely ， distributed database becomes a significant part of information management it avoids a lot of disadvantages of conventional centralized database ， and is applicable to many situations distributed database is a set of data which is the unity logic ， but in fact these data is located on different sites it is of high availability ， easy expansibility ， high concurrency ， high efficiency and etc in distributed system ， data redundancy is a method to improve the speed of query and the availability of system distributed query should shield the lower level details of data redundancy from end users ， distributed transaction should ensure data from disaccord this paper introduces basic conceptions of distributed database firstly ， discusses distributed transaction and concurrent control ， describes the development prototype mysql ’ s characters ， architecture and executive mechanism ， then shows the skeleton model of dpsql and exposes the strategy and algorithm of distributed query and distributed transaction ， at last analyzes the extra expenses and response delay of distributed processes this paper exposes the implementation mechanism of distributed query and distributed transaction emphatically distributed query uses the strategy of “ read one ， write a11 ” s0 in such a system if user ’ s requests are read - - only for the most part and the distribution of data redundancy is plausible ， efficiency is very high distributed transaction uses two - 。 phase commit protocol to ensure the consistency of global data ， which has less communication overhead
分散式查詢需要向用戶屏蔽數據冗餘分散的底層細節，分散式事務處理要保證全局數據完整，這都是傳統集中式數據庫不曾面臨的問題。本文首先介紹了分散式數據庫的概念性問題；接著討論了分散式事務涉及的定義及演算法；然後描述了dpsql的原型mysql的特性、結構和執行機制；繼而給出了dpsql的梗概模型，闡述了實現分散式查詢和分散式事務處理的策略及演算法；最後分析了進行分散式處理給系統增加的額外開銷和客戶端的額外響應延遲。本文重點描述了分散式查詢和分散式事務處理的實現機制，分散式查詢採用」讀一個寫全部」的本地優先策略，在讀操作頻繁的系統中，只要庫的分佈合理，單機效率極高，幾乎無任何附加開銷和延遲，而以整個系統的角度看來，吞吐量就更是優于單機系統；分散式事務處理採用兩階段提交協議，通信次數較少並能確實保證副本一致。
分享友人