Reduction of Power Consumption in Cloud Data Centers via Dynamic Migration of Virtual Machines

Abstract

Abstract-Cloud computing is the latest answer of technology to meet the computational requirements of users. The notable point in complicated computational works is energy consumption. The integration is one of the elements in the cloud system which can reduce the energy consumption and coordinate the software products. In this article, some solutions have been considered for determining the upper bound threshold of utilization for doing migration in order to reduce the power consumption. Also, it has been tried to use a solution to diagnose the overloaded host and eliminate it from the host as it can amend the efficiency as well as improve the host performance. All these will eventually reduce the power consumption. The results of evaluation show that the presented model has better efficiency in reducing the power consumption as well as decreasing the number of the migrations in comparison with the other models.Index Terms-Cloud computing, overloaded host, power consumption, migration. I. INTRODUCTIONmodel, similar to any other technology, has been created according to its own specific advantages while it involves some disadvantages. One of cloud computing disadvantages is producing a huge amount of CO2 due to the presence of very large data centers and high power consumption. The studies [2] regarding power consumption of server farms in 2005 show that the electric consumption by the wide world of servers (including cooling servers and their equipment) cost US $ 7.2. This study also shows that electricity consumption in that year, compared with consumption in 2000, has been doubled. However, the technology improvements have presented suitable hardware [3] including low-power processors, solid state drives and energy efficient monitors. A series of software solutions are optimally improving the energy efficiency. These two guidelines (hardware and software) should be evaluated as competitive supplements in order to incredibly increase their effects on the amount of energy consumption. The software solutions alongside software methods can improve the cloud computing Manuscript received September 12, 2014; revised November 17, 2014. Mona Arjmandi, Nik Mohammad Balouchzahi, Mahmood Fathy, and Ahmad Akbari are with the Iran University of Science and Technology, Tehran, Balouchzahi@iust.ac.ir, mahfathy@iust.ac.ir, Akbarii@iust.ac.ir).Kaamran Raahemifar is with the Ryerson University, Toronto, Canada (e-mail: Kraahemi@ee.rayerson.ca).efficiencies from the power consumption point of view. In this article one of the software factors, integration of virtual machines, will be evaluated. This factor reduces the server's power consumption. In this solution the low-load and overload servers will be diagnosed through a proper method and after migration of virtual machines, the low-load servers will be turned off. Moreover, those groups of the virtual machines which belong to the overload servers that ought to migrate are determined via the selective policy of the virtual machines. In the next step the selected machines wil...

Power Efficient Strassen’s Algorithm using AVX512 and OpenMP in a Multi-core Architecture

“…In addition, dynamic power consumption can be significantly decreased by reducing the power supply voltage to the defined level that provides the required performance as a dynamic performance scaling at the hardware level. Some techniques in [20][21] energy-efficient dynamic Virtual Machines (VM) consolidation are applied at the single-server or multi-server level for the distributed system.…”

“…Since it rounds the result only once, while separate multiplication and addition operations have two. Most compilers can perform auto-vectorization, and complierdirected auto-vectorization has strong limitations in the analysis and code transformation phases that prevent an efficient extraction of SIM D parallelism in real applications [20].…”