On the Isoefficiency Analysis of Parallel Computers with Linearly Scale-up Memory Space

等同效率用於分析線性擴展記憶空間之平行電腦的延展性之探討

林作俊(Cho-Chin Lin)

《宜蘭技術學報》 9期電機資訊專輯 (2002/12) Pp. 139-145

https://doi.org/10.6179/nit.eec.2002.09.24

摘要

研究顯示給予一個可延展平行電腦夠多的資料量，它的運算效率可以維持在一個常數。這樣的特性，對於在有限的時間內解決運算複雜且資料量龐大的問題提供了美好的前景。在[1，2]中，作者提出等同效率函數用以分析平行電腦的延展性。然而他們的分析中忽略了記憶體對延展性的影響。在此篇論文中，我們以等同效率函數為基礎，提出在有限的記憶資源中分析平行電腦延展性的方法。我們所獲得的結論如下:當平行電腦之每個運算模組的記憶體容量固定時且該平行電腦的執行效率必須維持在一個常數峙，此平行電腦之運算模組的數量不一定可以無限制的擴展。

關鍵字

延展性 ； 等同效率 ； 平行電腦 ； 記憶空間

並列摘要

A scalable parallel machine may maintain its efficiency constant by increasing the amount of input data. Thus, it has the opportunity to challenge the applications of large scale. In [1, 2], isoefficiency function has been proposed as a measure of scalability. However, their analysis ignored the amount of memory in a parallel machine. In this paper, the isoefficiency function of a parallel machine with memory constraint is proposed. Our result shows if the amount of memory of a parallel machine is linearly proportional to the number of processing nodes and the efficiency needs to be maintained at a desired level, then, the maximal number of processing nodes of the system may be bounded. This implies that the amount of available memory also affects the scalability of the system.

並列關鍵字

Scalability ； Isoefficiency ； Parallel Computer ； Memory Space