Virtual Machine Placement And Power Attack Detection In Multi-Tenants Cloud Data Centers

With the thriving development of cloud computing and big data industries,the number and scale of data center has expanded exponentially.As an information infrastructure,the data center plays an important role in offering computing,storage,and network services,but there exist some problems such as high energy consumption and low resource utilization.The virtualization technology is wildly applied in data center to improve the resource utilization since it can dynamically organize multiple resources,isolate the software from the specific hardware and gain a transparent and scalable computing system architecture.Besides,because of the dynamic changes in the virtual machine workload of different tenants,the data center tends to adopt the over-subscription technology to fully utilize the power resource for a significant operating cost savings.Although the over-subscription technology improves the power usage density,it increases the risks of power shortage,excessive current,and even circuit breaker trips during the peak load periods.The data center power attack may lead to server failures,circuit breaker trips and even power supply interruptions by maliciously increasing the power consumption of servers and other equipment thereby exceeding the limit of the power distribution system.As the widespread deployments of these technologies such as multi-tenancy,virtualization technology,containerization,and power over-subscription,the highdensity aggregation of data center servers and virtual machines has exacerbated the risk of data center power attacks.The malicious tenants usually leverage the controlled virtual machines to launch a power attack in multi-tenant cloud data centers.Therefore,it is also necessary to detect and control the malicious virtual machines in an earlier time,which can not only reduce the threat of power attack but also eliminate the serious consequences of power attack.In this dissertation,based on the studies of virtual machine placement,power attack detection and power attack protection,we conduct a research on the energy savings and security of multi-tenant cloud data center.The main contents of this dissertation are depicted as follows.(1)We propose a virtual machine placement strategy based on the genetic algorithm,and this strategy can carry out an effective placement for virtual machines which are calculated by a LSTM prediction model.The experiment result shows that our algorithm can greatly improve the resource utilization and reduce the energy consumption compared with the first fit and best fit algorithm,specifically,it reduces the resource fragmentation by 53.3% and 31.3% respectively,and it can also reduce the server deployment overhead by 6.9% and 10%.(2)We propose a power attack detection and positioning strategy which can accurately detect the power attack in data center and locate the malicious virtual machines.This strategy detects the potential power attack timely by comparing the realtime power consumption and resource utilization with a power attack recognition database based on the real historical power consumption and resource utilization data.And it further locates the malicious virtual machines by a fine-grained virtual machine power consumption estimation model and the power attack positioning algorithm.The experiment result shows that our algorithm can locate the malicious servers with 75% accuracy,and it can locate a malicious virtual machine with 100% accuracy if we accurately locate the malicious server.(3)We put forward a power attack protection strategy which can effectively manage and control the data center power consumption and further solve the data center power attack by combining DVFS and our adaptive multi-threshold energy efficient virtual machine migration algorithm—EEVMM.The experiment result shows that our strategy can save approximately 30% energy consumption without degrading the normal virtual machines performance and increase the difficulties in another malicious power attack.Power attack is one of the neoteric threats that the data center faces.In this dissertation,we propose an optimized virtual machine placement strategy to improve the data center resource utilization and reduce the energy consumption.However,the strategy may lead to power attack in multi-tenant data center because of the highdensity of virtual machines.Therefore,we also focus on the solutions to deal with the power attack to guarantee data center power security.And the experiment result shows that the strategies proposed in this dissertation can maximize the resource utilization and ensure the power security at the same time for data centers.