V. F. Nicola and A. Goyal. Modeling of correlated failures and community error recovery in multiversion software. IEEE Trans. Softw. Eng., 16(3):350--359, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is based on conditional probabilities. The second approach is based on the Beta Binomial distribution. The Beta Binomial distribution has been used previously to model correlated events such as the number of infectious diseases in a household [Griffiths1973] and failures in multi version software [Nicola1990]. 4.3.2 Datasets Used In our work we use two datasets. The first one is our availability measurements of Web Servers and the second one is measurements of Desktops [Bolosky2000] Web Servers. This dataset consists of our measurements of more than 100 web servers from September 2001 to December ....

....For P (4 up) there is only one path. 4.3.4 The Beta Binomial Distribution The second model we suggest for modeling the availability of correlated failures is based on the Beta Binomial distribution. The Beta Binomial distribution has been used to model other correlated events [Griffiths1973] [Nicola1990]. The distribution is similar to the model based on conditional probabilities in the sense that it uses average availability and a correlation level as parameters. The Beta Binomial distribution is computed by randomizing the parameter (p=failure probability) of the Binomial distribution. In ....

V. F. Nicola, A. Goyal "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software". IEEE Transaction on Software Engineering, Vol. 16 No.3, March 1990

....to be estimable in practice. We are unlikely to be able to estimate ) for any particular x, since this would require us to have a lot of independently developed programs that we could use to execute x. Thus the key variance term in (4. 6) will not be estimable for a new development, although [Nicola Goyal 1990] shows ways of estimating it given a sample of many developed versions of a program obtained in experiments. 11 For an interesting discussion of the scientific controversy that surrounded some of these early claims, see [Knight Leveson 1990] and the sources quoted there. Bev Littlewood, Peter ....

V. F. Nicola and A. Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software", IEEE Transactions on Software Engineering, 16 (3), pp.350-359, 1990.

....and simulation [8] and generalized stochastic Petri nets [13] model types. On the other side, the second approach based on the ground breaking work of Eckhardt and Lee [7] considers the precise meaning of the independence referred to the failure behavior of the diverse program versions [16] [19]. Recently in [11] 22] have been proposed NVP models that combine performance and dependability measures via the concept of performability. 2. Modeling approach: basic concepts Our approach to model NVP reliability is more general than previous ones as it combines information on the software ....

.... P ok ( jL = M = P tf ( jL = M = P ff ( jL = M = are given in (9) 8) 10) In general, we use the Lebesgue Stieltjes integral, thus covering both the discrete and continuous distribution functions H( 1 The last assumption, although usual in most software models [7] 16] [19] and software testing experiments is a simplification of real life, as it does not explicitly model the phenomena of failure clustering, that is the correlation of successive input states. 1e 10 1e 09 1e 08 1e 07 1e 06 1e 05 0.0001 0.001 0.01 0.1 1 1 5 9 13 17 21 n = 0:5 3 3 = ....

V. Nicola and A. Goyal. Modeling of correlated failures and community error recovery in multiversion software. IEEE Transaction on Software Engineering, 16(3):350--359, March 1990.

....are available several software development methodologies. The idea is that will vary not only from one x to another, but from one methodology to another. Other contributions in this class include modeling the intensity of coincident errors by a random variable with a Beta density function in [25] and the recovery block modeling considered in [30] which examines several methods for determining the intensity distribution: the use of the beta binomial distribution suggested in [25] and two new methods based on the pairwise correlation of modules. Safety critical real time computer ....

.... class include modeling the intensity of coincident errors by a random variable with a Beta density function in [25] and the recovery block modeling considered in [30] which examines several methods for determining the intensity distribution: the use of the beta binomial distribution suggested in [25], and two new methods based on the pairwise correlation of modules. Safety critical real time computer applications are characterized by stringent deadlines and high reliability requirements. In these terms, deadlines on the time that certain processes should be completed is performance ....

[Article contains additional citation context not shown here]

V.F.Nicola, A.Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software ", IEEE Trans. on Software Engineering, Vol.16, No.3, Mar 1990, pp. 350 -- 359.

....models [14] extended stochastic Petri net [15] and generalized stochastic Petri nets [16] model types. On the other side, the major goal of models that belong to the second class is the precise meaning of the independence referred to the failure behavior of the diverse program versions [17] [18], 19] The recently proposed performability models of the NVP in [20] and [21] place certain constraints on how properties affecting performance interact with those affecting dependability. The fault tolerant software system investigated here is for the real time mission critical ....

V.F.Nicola, A.Goyal, Modeling of Correlated Failures and Community Error Recovery in Multiversion Software, IEEE Trans. on Software Engineering, 16(3), 1990, 350 -- 359.

.... (x) will generally take different values for different inputs x. The situation when vary not only from one input to another, but from one development methodology to another is presented in [5] Other contributions in this class examine several methods for determining the intensity distribution [29], 23] In order to place certain constraints on how properties affecting performance interact with those affecting dependability, a unified measure called performability has been introduced [20] Until the recently proposed performability models of the NVP in [13] and [4] model based evaluation ....

V.F.Nicola and A.Goyal. Modeling of correlated failures and community error recovery in multiversion software. IEEE Trans. on Software Engineering, 16(3):350--359, March 1990.

....correlation between the failure behavior of redundant module versions. Consequently, Belli and Jedrzejowicz suggest in [2] to extend software reliability optimization approaches by taking failure correlations explicitly into account, for example based on the correlation model by Nicola and Goyal [10]. The present paper adopts this suggestion and elaborates some results on optimal redundancy distributions for module versions with correlated failures. As the basic system structure models, we have chosen Model 2 and Model 4 in [3] The authors also treat two other models, which, however, do ....

....refer to recovery blocks where the testing segments are assumed to be fault free. Compared to the models in [2] this is a simplification which lends itself as a base for the generalization to the more complicated correlated failures case. Our failure correlation model has been taken from [10]. In this article, the authors specialize a general probabilistic approach of describing coincident errors, developed by Eckhardt and Lee [5] to the particular case of beta distributed failure intensity functions. The co reliability of redundant versions of a software module is then ....

[Article contains additional citation context not shown here]

Nicola, V. F., Goyal, A., "Modeling of correlated failures and community error recovery in multiversion software", IEEE Trans. Software Eng., vol. SE-16, 350 -- 359 (1990).

....The work by Littlewood and Miller [18] indicates the need for research to objectively measure development processes in quest for less (in the best case negatively) correlated methodologies. To the best of our knowledge, little progress has been reported in this direction so far. Nicola and Goyal [21] suggested the Beta distribution as a suitable representation of the probability of failure of a randomly selected version on a randomly selected input, with the two model parameters to be determined from testing. These authors refer to the experiment by Knight an Leveson [16] to show the ....

V. F. Nicola and A. Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software", IEEE TSE, 16, pp. 350-359, 1990.

....used. The work by Littlewood and Miller [8] indicates the need for research to objectively measure development processes in quest for less (in the best case negatively) correlated processes. To the best of our knowledge, little progress has been reported in this direction so far. Nicola and Goyal [9] suggested the Beta distribution as a suitable representation of the probability of failure of a randomly selected version on a randomly selected demand, with the two model parameters to be determined from testing. These authors refer to the experiment by Knight and Leveson [10] to show the ....

V. F. Nicola and A. Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software", IEEE Transactions on Software Engineering, 16, pp. 350-359, 1990.

....controller computers. Several researchers proposed models for correlated faults in hardware or software. In [4] a simple model of correlated faults treated the case of two modules using linear correlation but did not include coincident faults induced by a common source. The author of [5] considered correlated faults in multiversion programming for fault tolerant software by using an appropriate intensity (or distribution) function. An example in [1] treated the problem of failure patterns in separately located computers to investigate the effects of a common environment on ....

V. F. Nicola and A. Goyal, "Modeling of correlated failures and community error recovery in multiversion software," IEEE Trans. on Software Eng., vol. SE-16, no. 3, pp. 350--359, March 1990.

....of the paper: element denotes one of the M version segments in a M by N (or MxN) system. stage denotes a 1 by N horizontal slice of a M by N system. version denotes a M by 1 vertical slice of a M by N system. Reliability of Community Error Recovery has been investigated theoretically [3, 11, 13] and experimentally [4] Nicola and Goyal [11] results indicate that the effectiveness of the technique is highly sensitive to, and easily degraded by errors in CER check points and interversion failure dependence, particularly when very high system reliabilities are targeted (e.g. 0.99999 or ....

....version segments in a M by N (or MxN) system. stage denotes a 1 by N horizontal slice of a M by N system. version denotes a M by 1 vertical slice of a M by N system. Reliability of Community Error Recovery has been investigated theoretically [3, 11, 13] and experimentally [4] Nicola and Goyal [11] results indicate that the effectiveness of the technique is highly sensitive to, and easily degraded by errors in CER check points and interversion failure dependence, particularly when very high system reliabilities are targeted (e.g. 0.99999 or more) CER is more robust at lower target system ....

[Article contains additional citation context not shown here]

V.F. Nicola, and Ambuj Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multi-version Software," IEEE Trans. Soft. Eng., Vol. 16(3), pp, 1990.

....be computed for any time, which will suffice for performing the cost effectiveness analysis. Recent experimental results seem to suggest that the failures of different software versions may not be independent [19] A few models have also been proposed to model correlated failures in NVP software[10, 24]. However, these models define reliability of a software as a function of input values to the software. That is, reliability is defined as the probability that a given software will fail at a given input. This definition is not suitable for our purposes, since we want to compute the time required ....

....therefore R 2 (t) and R 3 (t) must be parametric with respect to correlation. We therefore need a measure of correlation between softwares. Correlation modeling turns out to be a hard problem. Some effort has been done to model the reliability of a software system with correlated failures [10, 24]. These models typically try to predict the probability of failure, for a given input. Such an approach is not suitable for our work as we need the reliability as a function of time. We used the statistical method of determining the correlation coefficient [28] ae = E(XY ) GammaE(X)E(Y ) oe X ....

V. F. Nicola and A. Goyal, "Modeling of correlated failures and community error recovery in multiversion software", IEEE Tran. on Software Engg., Vol. 16, no. 3, March 1990, 350-359.

.... Software (NVS) systems, through the development and use of design diversity [1] Such systems are normally operated on an N Version Executive (NVX) environment [2] Since the idea of NVP was first proposed in [3] numerous papers were published in either modeling NVS systems [4] 5] 6] 7] 8] [9] [10] or providing empirical studies of NVS systems [11] 12] 13] 14] 15] 16] 17] 18] The effectiveness of the NVP approach, however, has remained highly controversial and inconclusive. Nevertheless, most researchers and practitioners agree that a high degree of version independency and a ....

V.F. Nicola and A. Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multiversion Software," IEEE Transactions on Software Engineering, vol. SE-16, no. 3, pp. 350359, March 1990.

No context found.

V. F. Nicola and A. Goyal. Modeling of correlated failures and community error recovery in multiversion software. IEEE Trans. Softw. Eng., 16(3):350--359, 1990.

No context found.

V.F. Nicola, and Ambuj Goyal, "Modeling of Correlated Failures and Community Error Recovery in Multi-version Software," IEEE Trans. Soft. Eng., Vol. 16(3), pp, 350-359, 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC