國立臺灣大學電機工程學系暨研究所郭斯彥2006-07-252018-07-062006-07-252018-07-061999-06-30http://ntur.lib.ntu.edu.tw//handle/246246/7737隨著計算機系統的發展，計算機系 統已廣泛的應用於各行各業。包括一些 具有高危險性或重要的工作，例如核電 廠控制系統，手術室的醫療監視系統， 以及飛行控制系統。但是計算機本身是 由硬體電子零件以及軟體邏輯程序的組 合，任何一個部份做出預料之外的反 應，都有可能導致整個計算機系統失常 或者停頓。對於上述之高危險性與重要 的工作，計算機系統的故障會造成鉅大 金錢與生命的損失。尤其是用於軍用或 是民間飛行控制的電腦，由於在飛行過 程中飛行控制系統必須持續監控飛機的 狀況，並且即時做出正確反應，如果飛 行控制系統的某一部份發生故障，將會 造成極大災難。為了確保飛行安全，飛 行控制系統必須具備極高的可靠度，一 般來說，在系統運作過程中每小時發生 錯誤的機率必須小於10-9 以下。 近三十年來，關於『如何使計算機 系統更可靠』這方面的研究，有了重大 的發展。在早期的系統中，增進系統可 靠度的方式即是經由不斷的測試、修改 而建構出一個接近零缺點的系統。然 而，在系統建構的過程中，需要付出相 當龐大的人力、資源以及冗長的研發時 間。另一方面，工程人員可以藉由各種 備份技術，讓計算機系統本身具備容錯 之能力。依據設計理念的不同，容錯計 算機系統不只能容忍系統部件意外故障 所造成的錯誤，更能延伸到容忍系統設 計上的缺失，與人員操作不當造成的錯 誤。在近二十年中，由於微處理器的發 展，計算機系統的體積、重量、耗電量 與成本皆大幅縮減，使得硬體備份已確 實可行。利用這些技術，容錯計算機系 統將可以兼顧功能性與可靠度，同時將 成本壓低至實用的階段。本計畫將利用 微處理器技術，研究並實際完成一個能 應用在飛行控制系統中之容錯計算機雛 型(Prototype)。Computer systems are widely used in variety of tasks. For critical missions such as the medical system in the operation room, the aviation control system, or the nuclear reactor control system, malfunction of the computer system may cause great casualty. In the past three decades, one can see the significant growth in this issue: how to make the computer system more dependable ? One way to achieve higher dependability is to make the computer system fault-free; that is, to build a near perfect physical system, and execute a near perfect procedure. However, it is a dependabilityand- resources trade-off. As the computer systems are getting more and more complex, it is more and more difficult to design, build, and verify the system that can per-form perfect operation at any circumstance throughout the system's life time; and the cost to develop and build a near perfect system is increasing exponentially as the system scale and complexity increase. Another way to achieve higher dependability is to make the system fault-tolerant. A fault-tolerant system is a system that will keep operating normally and give the correct responses and answers when one or more of the components fail to perform it's normal task. A more general concept of fault-tolerance also includes man made fault, such as the human mistakes committed during hardware or software design and implementation, or bugs in short term. Another possible human mistake is the operating error due to the improper design of human computer interface, or the lack of training for operators. To achieve fault-tolerance, redundant hardware is required. It is a big problem in the early era of computer development while the computers are in the size of a refrigerator, and they are costly. With the advent of microprocessors, the volume, weight, power and cost associated with redundant component decreases dramatically. With microprocessors, faulttolerant systems can be built highly dependable and affordable.application/pdf42304 bytesapplication/pdfzh-TW國立臺灣大學電機工程學系暨研究所容錯計算機三組態備份容錯微處理器表決器Fault TolerantTMRMicroprocessorFPGAVoterreporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/7737/1/882623D002012.pdf