Study On Virtual Machine Scheduling Algorithm Based On Hybrid Memory In Embedded Systems

The hardware and software development of embedded intelligent system makes the foundation of mobile virtualization technology. Mobile virtualization provides a secure and reliable mobile ecosystem in many aspects, such as personal privacy, corporate and public security. In addition, mobile virtualization reduces the cost of the hardware and software and shortens the software development cycle. However, mobile virtualization posts new challenges on real-time scheduling and hardware resources due to multiple virtual machines sharing resources in the virtual environment. Fortunately, NVRAM(Non-Volatile Random Access Memory) provides opportunities to extend the system memory occupies and enhance the non-volatile of the system, it can also solve the problem of the traditional DRAM storage technology such as scalability and energy consumption with its attractive features, like high density, low standby power and so on. However, the long write latency, limited life and high energy consumption of NVRAM may slow down the performance of write-intensive tasks and time tasks.Therefore, in this work, we focus on the problem of limited resources of embedded system based on KVM virtualization technology, introduce the NVRAM/DRAM hybrid memory as the embedded device main memory, and put forward a task allocation algorithm in virtual machine. Then based on the characteristics of the embedded terminal users, we propose an unfair scheduling algorithm in mobile virtualization system. Finally we design a test to verify the effectiveness of the proposed algorithm. This work makes the following contributions:First, the presence of multiple virtual machines in virtual environment increases the read, write and updated operations of main memory. And too many write operations of NVRAM not only shorts the life of NVRAM, but also brings a certain negative impact on performance of a system. So, in this paper, we propose a task allocation algorithm to resolve this problem. So as to reduce the system energy consumption, prolong the life of nonvolatile memory, and increase the utilization ratio of hardware, the task allocation algorithm based on the advantages and disadvantages of NVRAM and DRAM memory, distributes the read-intensive tasks running in NVRAM and distributes the write-intensive tasks running in DRAM.Second, the increased number of virtual machine and the long write delay of NVRAM post new challenges of real-time tasks scheduling to mobile virtualization. So, in this paper, we propose an unfair virtual machine scheduling algorithm to resolve this problem. The unfair scheduling algorithm promotes the priority of the foreground virtual machine according to the users using habit of embedded devices, and evenly distributes virtual CPUs to physical CPUs to guarantee the real-time task can be scheduled, and improve the user experience.Third, to verify the feasibility of the proposed algorithm of this paper, we implement a KVM-based mobile virtualization platform in Linux system. The experimental results based on various pseudo-randomly selected task sets show that the proposed algorithms can improve the system performance at least 10% by tasks completion time in average, and significantly increase real-time tasks' completion rate to about 100% compare to the default CFS scheduling method.