1336
World Abstracts on Microelectromcs and Reliability
and post-service delivery stage into one single model, similar to the structure of a total m a n u f a c t u n n g quality management system The design specification was described by means o f data flow diagrams (DFD) The model can be used as the base for further development towards a computer integrated service quality management information system
A multi-period design model for survivable network architecture selection for SONET interoffice networks. TSONG-Ho WE, RlCrtMU3 H CARDWELL and MiKE BOYDEN IEEE Trans Rehab 40(4), 417 (October 1991) Future survivable S O N E T interoffice networks probably will use a mix of diverse protection and ring network arch3tectures to provide affordable network survivablhty This paper proposes a multi-period design model to select a best mix o f S O N E T survivable network architectures These include 1) point-top o m t / h u b b m g architecture with diverse protection, and 2) ring architecture The algorithm can be used to study growth strategies, for networks incorporating the n n g architecture, that mtmmize the network evolution cost Numerical examples show that the algorithm can generate a good solution quickly, using a VAX 6000-420 computer, for most practical applications A case study suggests that a particular well accepted engineering rule can work well This engineering rule is_ use direct point-to-point systems in conjunction with self-healing n n g s and hubbmg/dwerse protection architectures only when a few demand pairs have appreciably higher demand requirements than other demand pairs_ When demand is fairly even, then use selfhealing n n g s or hubblng/dlverse protection without direct point-to-point systems
An expert aid to fault diagnosis and trouble-shooting in satellite communications. M B GREEN and P R SIMs Electron. Commun Engng J , 252 (December 1991) As a result of the increasing complexity of commumcation systems there is a need for a level of system knowledge a m o n g staff at each site that cannot easily be met If some percentage of this knowledge can be formahsed as an expert system, it can be used as a support tool and "expert adwsor" for system operators. The paper presents the Satelhte Communications Network Expert (SANE), a knowledge-based aid designed to provide engineering support for fault diagnosis_ The SaNE is composed o f two elements a satcom network model, which simulates the structure and functionahty of a system based loosely upon a large military satcom network, and a diagnostic component, which uses knowledge- and model-based reasoning techniques to analyse system anomalies and diagnose possible causes for the alarms such networks generate The development cycle is described, emphasislng lessons learnt during development and testing and the advantages and disadvantages of the techniques applied The goal of the SaNE project IS p n m a r dy commercial acceptance rather than m n o v a t m n It is the aim of this paper to illustrate how novel concepts can be implemented in a practical system without compromising this goal
Reconflgnrable fault tolerant networks for fast packet switching. SHIH-CHIAN YANG and JOHN A SILVESTER IEEE Trans, Rehab 40(4), 474 (October 1991) Fast packet switching is intrinsic to many apphcatlons that use fiber opUc technology As the technology is pushed to improve speed, network rehabfllty becomes a key issue. The design constraints resulting from the use of fast packet switching that impact fault-tolerant network design are carefully studied. A reconfigurable network with high link redundancy IS then proposed An abstract replacement model that characterizes the proposed reconfigurable network is presented Network fault tolerance problems are transformed into well known assignment problems The abstract model
prowdes an effectwe tool to design and analyze practical fault tolerant networks for fast packet switching More reliable networks with less hardware overhead can be designed w~th the techniques shown in this paper Two practical designs based on feasible technology are presented An appreciable rehablhty improvement is achieved and full bandwidth is maintained up to a tolerable level of failures, with relatively few spare switches
Reliability enhancement by time and space redundancy in multistage interconnect-ion networks. V P KUMAR and S J WANG IEEE Trans Rehab 40(4), 461 (October 1991) Multistage Interconnectlon Networks (MINs) offer an attractive way of Implementing fast packet switches In communication networks With the throughput requirement of the packet switches exceeding several glgabits per second, it becomes imperative to make them fault tolerant Fault tolerance in M I N s can be achieved by space redundancy or time redundancy When space redundancy is used, multiple paths are created between input-output pairs, so that the full access property (the abdity to reach any output from any input in exactly one pass) is preserved m the presence of some faults W h e n time redundancy is used, multiple passes are made through the M I N to reach the desired destination Dynamic Full Access (DFA), or the ability to route from any input to any output in a fimte number of passes through the network, has been suggested as a criterion for the fault tolerance of M I N s A weakness of D F A as apphed to shuffle-exchange and other unique path M I N s is that the switches in the f r s t stage and in the last stage comprise a hard core, ~ e , faults in them cannot be tolerated This paper presents the D F A properties of fault tolerant M I N s which have multiple connections to the inputs and outputs, and thus potentmlly no hardcore When full access is lost in these fault tolerant M I N s due to multiple faults, but D F A exists, multiple pass routing could be utilized to achieve graceful degradation Some sufficiency c o n d m o n s for the existence of D F A in a broad class of fault tolerant M I N s are derived The rehablhty of four multiple path MINs under D F A is studied The metrics used are the probabdlty of existence of DFA, and the mean time to failure The study found that one particular network (the MD-Omega) which uses a mlnlmurn a m o u n t of hardware redundancy to provide two connections from each source to the M I N and to each destination from the MIN, shows the most gain m rehablhty when time redundancy IS used The M D - O m e g a network has a 2 x 2 switch as its basic element, but is almost as reliable as another fault tolerant MIN, the ASEN, which uses a 3 × 3 element, when multiple pass routing is used Thus, for the same size VLSI building block chip, it has the potential to have wider data buses and, therefore, higher performance under fault-free condltions On Murphy's yield integral. CHARLES H STAPPER IEEE Tram Semtcond Mfg 4(4), 294 (November 1991) Recent papers on integrated ctrcmt yield modeling have reactivated interest in the ancient technique for estimating integrated circuit m a n u f a c t u n n g yields by means of a compounding technique originally described by Murphy It is assumed in these papers that the use of a Gausslan distribution as a model for the variation in defect densities requires the use of a triangular approximation However, such an approximation is not necessary It IS shown in this paper that an exact solution is posmble The resulting yield formula has actually been used in the industry for a short ume, but has been superseded since by use of the negative binomial yield model which Is easier to manipulate and in better agreement with actual data