Derivation of a Waiting-Time Factor for a Multiple-Bank Memory
Journal of the ACM (JACM)
A Survey of Analytical Time-Sharing Models
ACM Computing Surveys (CSUR)
Markovian models for performance evaluation of multiprocessor, multimemory computer systems
Markovian models for performance evaluation of multiprocessor, multimemory computer systems
Introduction to Probability Models, Ninth Edition
Introduction to Probability Models, Ninth Edition
| Hi-index | 0.00 |
In a multiprocessors configuration, a memory conflict occurs whenever two or more processors or I/O, s attempt to access the same memory unit simultaneously. The overall effect of these conflicts are lose of the system efficiency. A multiprocessor activity during execution of a program has been considered as a stochostic process which moves through a countable set of states. The process decides where to go next by a random mechanism which depends only on the current state and not on the previous history or even on the time. This kind of process is called Markov process. The transition from any state is possible only to the neighbouring states due to these conditions that the arrivale of request rate from processors and the memory communication rate follow a poisson distribution. Under these assumations a multi-server waiting line model has been developed for a multi-processors organisation. The basic motivation for this investigation will be the search to find a set of system parameters to measure the multiprocessors. performance with multi-memory nuits.The Makrovian waiting line model has been emplayed and obtained a general solution to estimate the effect of memory conflict on the average number of waiting units in the system, idle and busy periods for memory unit, system efficiency, the stretching factor of program on bases of "First Com, First Served" descipline.The simulation programs written by Basic to implement the developed waiting line model are general enough so that a system configuration of any size can be studied.