Sharif Digital Repository

In this thesis, three different mathematical models named Markov Reward Model (MRM), Stochastic Reward Net (SRN) and Stochastic Activity Network (SAN) are used to model and evaluate the performability of grid computing environments consisting of many grid resources. The proposed models consider the arriving and servicing process of grid tasks inside a resource together with the failure-repair behavior of processors of the resource. Since the proposed MRM cannot be extended to model a grid environment with some realistic assumptions, we switch to use SRNs in modeling a single grid resource with more number of processors. Although the proposed SRN models for a single grid resource can... 

Reconfigurable devices are a popular platform for fast prototyping of digital system due to having high performance of hardware implementation along with flexibility of software. However, reconfigurable devices suffer from area, performance and dependability gaps in comparison with their Application Specific Integrated Circuit (ASIC) counterparts, which greatly limits their application.
The dependability gap originates from the sensitivty of configuration memory to soft errors. When a reconfigurable device configuration memory is affected by soft errors, their configuration will be invalid until reconfigured. Since the routing fabric is the origion of over 80% of soft errors in... 

Storage subsystem is one of the most important components of a computer system. As more valuable information such as bank transactions are being stored in storage subsystems,occurrence of an unrecoverable error can lead to a much bigger problem. A storage subsystem can be divided into software and hardware part. Previous works focused on fault injection in the higher layers of the storage stack and didn’t consider fault injection at the device driver level. In this thesis, the dependability of storage stack is evaluated using fault injection at the device driver layer. Various hardware and software faults that can occur in hard disks and solid state drives are studied and dependability of... 

Reliability is an important concern in safety-critical applications. Hardware and software redundancies are used to include reliability in such systems. Software approaches benefit from high flexibility and low cost. Moreover, they do not impose any hardware modifications to the design. The most important challenges of software approaches are performance and code size overheads. Some of software reliability approaches duplicate all of the program’s instructions which impose high performance overhead to the designs. Some other approaches only apply the redaduncies to critical instructions in the programs. While these approaches impose the the performance overhead but they suffer from fault... 

Nowdays, embedded systems need high dependability and performance and low cost. Use of multicore processors cause increase performance, and can lead to increase dependability with its inherent redundancy. Also mixed critical leads to decrease cost and increase dependability, because designer just considers dependability techniques for high critical tasks. A lot of work to increase dependability of real-time multicore systems with mixed-criticality has been done. In this study, a comprehensive classification of the approaches that has been done in this area is provided. One of the issues that is expressed in muticore, is shared resources. In multicore platforms some resources like memory and... 

When multicore processors entered in industry and parallel processing became popular, most challenges have started. Some of the challenges include binding and time sharing of an application with deadline meating constraint. Another important challenge is that, these kinds of processors become more susceptible to transient faults. Among solutions to the mentioned challenges, virtualization which was one of the well known ones becomes applicable again because of some of its advantages such as flexibility, high abstraction and low cost implementation. But this technology has some drawbacks too. It can make system complex and leads to memory and performance overhead. Constructing a reliable... 

Considering current problems in performance evaluation of the Cloud and its management policies and strategies, this thesis presents different approaches for simultaneous evaluation of the performance, dependability, and power consumption of Infrastructure-as-a-Service (IaaS) clouds. The proposed approaches use Stochastic Reward Nets (SRNs) and Stochastic Activity Networks (SANs) to model and evaluate performability and power consumption of cloud systems. In all of the proposed approaches, the cloud resource management system is power-aware. It means that idle Physical Machines (PMs), network equipment, and supporting subsystems are kept in powered-off or standby modes, as much as possible,...