Cloud Workload Allocation Approaches for Quality of Service Guarantee and Cybersecurity Risk Management

It has become a dominant trend in industry to adopt cloud computing --thanks to its unique advantages in flexibility, scalability, elasticity and cost efficiency -- for providing online cloud services over the Internet using large-scale data centers. In the meantime, the relentless increase in demand for affordable and high-quality cloud-based services, for individuals and businesses, has led to tremendously high power consumption and operating expense and thus has posed pressing challenges on cloud service providers in finding efficient resource allocation policies. Allowing several services or Virtual Machines (VMs) to commonly share the cloud\u27s infrastructure enables cloud providers to optimize resource usage, power consumption, and operating expense. However, servers sharing among users and VMs causes performance degradation and results in cybersecurity risks. Consequently, how to develop efficient and effective resource management policies to make the appropriate decisions to optimize the trade-offs among resource usage, service quality, and cybersecurity loss plays a vital role in the sustainable future of cloud computing. In this dissertation, we focus on cloud workload allocation problems for resource optimization subject to Quality of Service (QoS) guarantee and cybersecurity risk constraints. To facilitate our research, we first develop a cloud computing prototype that we utilize to empirically validate the performance of different proposed cloud resource management schemes under a close to practical, but also isolated and well-controlled, environment. We then focus our research on the resource management policies for real-time cloud services with QoS guarantee. Based on queuing model with reneging, we establish and formally prove a series of fundamental principles, between service timing characteristics and their resource demands, and based on which we develop several novel resource management algorithms that statically guarantee the QoS requirements for cloud users. We then study the problem of mitigating cybersecurity risk and loss in cloud data centers via cloud resource management. We employ game theory to model the VM-to-VM interdependent cybersecurity risks in cloud clusters. We then conduct a thorough analysis based on our game-theory-based model and develop several algorithms for cybersecurity risk management. Specifically, we start our cybersecurity research from a simple case with only two types of VMs and next extend it to a more general case with an arbitrary number of VM types. Our intensive numerical and experimental results show that our proposed algorithms can significantly outperform the existing methodologies for large-scale cloud data centers in terms of resource usage, cybersecurity loss, and computational effectiveness