Modeling And Optimization Of Fault Self-recovery Time In DRP System

Fault self-recovery time (FSRT) is used to evaluate the reliability of the DRP (Distributed Redundant Protocol) system, which is affected by five parameters: macro cycle, offset of ring-check, limit of ring-check, offset of link-check and limit of link-check. The optimization problem is how to set the five parameters, in order to make the FSRT as short as possible.To analyze the FSRT, the message scheduling model and the fault self-recovery model are set up at the message level and process level, respectively. The message scheduling model includes fault detecting model and fault processing model, while the fault self-recovery model includes network self-establishing process and fault self-recovery process.Based on the above models, FSRT is analyzed in detail to get its range, whose lower and upper limit are the link failure processing time of the ring, and the sum of macro cycle and the link failure processing time of the ring, respectively. Then, the upper limit of FSRT is chosen to be objective function, and constraint conditions are presented according to the DRP's schedule principle and application backgrounds.During optimization, the five parameters are adjusted according to the theoretical results and experimental results. There are two innovations as follows.(1)A new method of division of work in fault detection is put forward. When the ring is pre-networking, ring-check message is used to detect fault, while when it's networking, the link-check message is used. Ring-check message and link-check message divide the work by the ring's condition, preparing for the time slices' sharing.(2)A method of sharing time slices is put forward when allocating macro cycle. The sharing of time slices, makes the limit of ring-check not occupy a separate time slice of macro cycle any more, greatly reducing the macro cycle. Simultaneously, FSRT and its upper limit are reduced.Using the above innovations, five parameters are set properly to optimize FSRT gradually in experiments, obtaining remarkable results.