Re: Software Reliability (10^-9 and all that)



Date view Thread view Subject view Author view

Pete Mellor (pm(at)csr.city.ac.uk)
Thu, 19 Jun 1997 22:57:11 +0100 (BST)


Difficult one, Jon! :-) A sensible approach might be something like the following:- - Avionics systems are generally modularised, and so individual software modules could be examined separately. These modules often run in separate bits of hardware. Example: A320 EFCS consists of 5 computers, each dual-channel with diverse software in the two channels. There are two computers of one type, and three of another. - It should be possible to establish the reliability of each module on the basis of extensive trials under simulation. - The resulting software reliability figures, and probabilities of common mode failure, could be used in the calculation of the overall system reliability, as the hardware figures currently are used. - Initially, a conservative estimate based on comparable software modules could be used (based on experience of the failure rates of similarly developed software from previous systems). - After experience with the current model, the figures could be updated to take account of in-service behaviour. All of this would require data to be recorded, broken down to module level. I think someone knows what these data are, but they ain't letting on! This is something that could be improved by a change to the regulations! Pete Mellor, CSR, City University, London <p.mellor(at)csr.city.ac.uk> ------------------------------------------------------------------- On Thu, 19 Jun 1997, Jon Davies wrote: > Peter Mellor said: > > [snip, hack, saw] > > There would seem to be a good case for having AMJ 25.1309 impose a claim > > limit on the reliability of software in such systems. > > and what would a sensible limit be? > > Cheers, > Jon > -- > Jon Davies, Senior Design Engineer > R&D, GEC ALSTHOM Signalling Limited, PO Box 146, MANCHESTER, UK. M60 1AX. > Tel: +44 161 875 2082 - Fax: +44 161 875 2085 - Telex: 667152 GALMCR G > Email: jdavies(at)iee.org >


Date view Thread view Subject view Author view