Research On Design Optimization Of Reliability-aware Real-time Tasks

Nowadays,embedded systems are deployed in different kinds of fields,along with the rapid development of computer technology.Especially in fields such as: aerospace,military industry and medical & health care.The architecture of embedded systems also transformed from single processing system,which is simple,to large scale distributed system.However,some of them are real sensitive,so it is very important for researchers to optimize the reliability and the security of relevant systems.Generally,an embedded system increases its reliability by using the methods of fault detection or fault toleration,on the other hand harden its security by introducing technology of encryption/decryption.While these harden techniques introduce not only the increasing of reliability and security,but also extra overheads.Sometimes,harden techniques may spend far more overheads than the embedded system itself.So it`s real necessary to find out some approaches which could decrease the overhead and,at the same time,guarantee reliability and security for embedded systems.In this thesis,we are interested in researching tasks which deployed on embedded system.We are going to harden the application`s reliability by using fault toleration technology.Also,we would like to analyze conditions how fault toleration technology effect during the scheduling.Our research optimizes whole embedded system on energy consumption,execution duration and vulnerability,and also explore the approaches solving problems above.Finally,we verify these approaches with simulation experiment.These are the contributions of this thesis:(1)Provide a design optimization approach which minimize energy consumption of periodic tasks running on distributed embedded system.Adopt an ‘imperfect fault tolerant' scheme based on multi-replicas technology.Analyze fault tolerant module then formulate reliability and energy consumption.Explore optimum mapping scheme by using simulated annealing algorithm.Finally,design experiments to prove our approach.(2)Provide a design optimization approach which minimize execution duration of tasks which have message constraints and running on distributed embedded system.Adopt multi-replicas and re-execution technologies.Analyze modules of how fault tolerant schemes make impacts on scheduling and communication.The only constraint in this chapter is the quantity of each single task.Our target is to explore a scheme of fault tolerant choosing which minimize execution duration by using genetic algorithm.Finally,design experiments to prove the scheme.(3)Provide a design optimization approach which minimize either execution duration or vulnerability of tasks which have message constraints and running on distributed embedded system.Adopt multi-replicas and re-execution technologies and analyze impacting modules.Our target is to explore a scheme of fault tolerant choosing which minimize either execution duration or vulnerability by using NSGA-II.Finally,design experiments to prove the scheme.