登入選單

返回Google圖書搜尋

Effects of failures on performance of gracefully degradable systems

Stanford University Stanford Electronics Laboratories. Digital Systems LaboratoryJ. Losq

出版	Digital Systems Laboratory, Center for Reliable Computing, Stanford University, 1976
URL	http://books.google.com.hk/books?id=r-oEAAAAIAAJ&hl=&source=gbs_api

註釋

The recent development of multiprocessor systems that offer resistance to faults by gracefully degrading after a failure opens vast new ranges of applications for fault tolerance and high reliability. The paper presents a general model for the evaluation of such systems. It takes into account the internal structure of the hardware, the characteristics of the various detection mechanisms, the unreliability of the software and even the type of applications these systems for which these systems are used. It provides many measures of the systems' performance such as: availability, meantime between crashes, average processing power and proportion of time spent in degraded mode. System optimization gives the best values for the number of processors, memories ..., and shows the trade-offs between hardware and software fault-detection mechanisms. The model is illustrated by a concrete example. (Author).