In future distributed computing systems, data-arrival rates, fault types, resource availability and user preferences among performance and dependability objectives, all may vary dynamically during operation. The report describes architectural approaches and techniques for building computer systems that dynamically adapt their structure to changing operating conditions and user preferences. The concepts of stability management and system identification from modern control theory are given digital-system interpretations. These are reflective control and real-time fault diagnosis. Several cases were studied and demonstrated, including adaptive distributed recovery blocks, and adaptive distributed thread maintenance. A formal scheme was developed for adaptive, hybrid byzantine-backup fault tolerance.

@techreport{csl-95-02,
    AUTHOR = {Jack Goldberg and Ira Greenberg and Raymond Clark and {E.} Douglas Jensen and Kane Kim and and Douglas {M.} Wells},
    TITLE = {Adaptive Fault-Resistant Systems},
    YEAR = {1995},
    MONTH = {jan},
    URL = {http://www.csl.sri.com/papers/sri-csl-95-02/}
}

• sri-csl-95-02.pdf
• sri-csl-95-02.ps