Abstract:
Google is currently the most influential Web search engine. It uses more than 10 thousand cheap pcs to construct a high-performance, ultra-large storage capacity, stable, and practical giant Linux cluster. This article analyzes the logic and physical construction methods, reliability, scalability, availability, and parallelism of the Google cluster system from the perspective of the computer system structure. This article focuses on the logic structure and physical structure of the Google cluster, the implementation of distributed file systems and ultra-large-capacity storage. According to the analysis in this paper, the high-availability and high-performance cluster method based on the characteristics of Google clusters for Web search needs is a success example of parallel machine design and development, this strict cost-effective design method is worth learning.
For the full text of PDF, see the link on the following webpage:
Http://hums.ccnu.edu.cn/teachers/yuyijiao.htm
Other references:
Http://en.wikipedia.org/wiki/Google_platform