TY - JOUR
T1 - A model for availability analysis of distributed software/hardware systems
AU - Lai, C. D.
AU - Xie, M.
AU - Poh, K. L.
AU - Dai, Y. S.
AU - Yang, P.
PY - 2002/4/15
Y1 - 2002/4/15
N2 - System availability is a major performance concern in distributed systems design and analysis. A typical kind of application on distributed systems has a homogeneously distributed software/hardware structure. That is, identical copies of distributed application software run on the same type of computers. In this paper, the system availability for this type of system is studied. Such a study is useful when studying optimal testing time or testing resource allocation. We consider both the case of simple two-host system, and also the more general case of multi-host system. A Markov model is developed and equations are derived to obtain the steady-state availability. Both software and hardware failures are considered, assuming that software faults are constantly being identified and removed upon a failure. Although a specific model for software reliability is used for illustration, the approach is a general one. Comparisons show that system availability changes in a similar way to single-host based software/hardware systems. Sensitivity analysis is also presented. In addition, the assumptions used in this paper are discussed. © 2002 Elsevier Science B.V. All rights reserved.
AB - System availability is a major performance concern in distributed systems design and analysis. A typical kind of application on distributed systems has a homogeneously distributed software/hardware structure. That is, identical copies of distributed application software run on the same type of computers. In this paper, the system availability for this type of system is studied. Such a study is useful when studying optimal testing time or testing resource allocation. We consider both the case of simple two-host system, and also the more general case of multi-host system. A Markov model is developed and equations are derived to obtain the steady-state availability. Both software and hardware failures are considered, assuming that software faults are constantly being identified and removed upon a failure. Although a specific model for software reliability is used for illustration, the approach is a general one. Comparisons show that system availability changes in a similar way to single-host based software/hardware systems. Sensitivity analysis is also presented. In addition, the assumptions used in this paper are discussed. © 2002 Elsevier Science B.V. All rights reserved.
KW - Clustered system
KW - Combined software/hardware
KW - Distributed system
KW - Markov model
KW - Software/hardware reliability
KW - System availability
UR - http://www.scopus.com/inward/record.url?scp=0037090799&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-0037090799&origin=recordpage
U2 - 10.1016/S0950-5849(02)00007-1
DO - 10.1016/S0950-5849(02)00007-1
M3 - RGC 21 - Publication in refereed journal
SN - 0950-5849
VL - 44
SP - 343
EP - 350
JO - Information and Software Technology
JF - Information and Software Technology
IS - 6
ER -