07-11-2016, 10:20 AM
[attachment=74696]
Abstract: Cloud computing emerged as a new computing paradigm for on-demand provisioning of IT services over the Internet. One challenging research issues in cloud environment it to develop any energy efficient and resource-aware algorithm for virtual machine placement (VMP) problem. If the problem is not dealt properly, it effects the performance of cloud data centers in terms of resource utilization and power consumption. In this paper, we propose a new VM placement for cloud environment, to address VMP problem. The objective of the proposed algorithm is maximize resource utilization and minimize the number of active physical machines. We adopt a clever strategy to assign ranks to VMs and placing VMs based on their ranks. We simulate the proposed algorithm as well as exiting algorithm. The simulation results show that the performance of the proposed is better than the existing algorithm.
1. Introduction
Cloud computing has emerged as a new computing paradigm which enables to provide computing resources to the cloud users over the Internet [1-3]. The computing resources are pooled at one place called datacenter and these centers are managed by an organization called cloud service providers. The service providers offer three types of services to the users such asInfrastructure as a Service (IaaS), Software as a Service, and Platform as a Service [4-6]. In IaaS, computing resources such as CPU, RAM, Storage and Network are offered as a service and their use is expected to abide by service level agreements [4]. Resources are shared among the users through virtualization. Virtual machines (VMs) are created over the physical machines (PMs), to execute the user's applications. One of the challenging problems in virtualization is to utilize the computing resources efficiently and minimize the power consumption. The above problem can be solved by employing an efficient virtual machine placement algorithm. The job of the placement algorithm is to place the pool of VMs on the PMs in such way that computing resources are efficiently utilized and also minimize the number of active PMs.
Numerous algorithms for VM placement have been proposed in the recent years [9-15]. However, most of the algorithms ignore the balanced utilization of resources which leads to higher resource wastage and also increase the number of active PMs. For example, PMs with d-dimensional resources, if utilization of all the resources are not balanced then it may happens that a newly arrived VM cannot be accommodated in any of the active single PM.As a consequence, increase in resource wastage of the existing active PMs and also a new PM needs to be activated in order to accommodate the newly arrived VM.This is due to the lack of an intelligent decision making while selecting VMs for placement. To demonstrate this, consider an example, as shown in Fig. 1.Virtual machines, VM1 and VM2 are placed in PM1, VM3 cannot be accommodated in PM1 and it is placed in PM2. VM4 neither housed in PM1 nor in PM2 and so it is placed in PM4. Due to improper consolidation of VMs, overall resources of the PMs are unused and also it leads to increase in number of active VMs. whereas, as shown in Fig. 2, due to proper consolidation of VMs, neither resources are wasted nor increase in number of active PMs. Therefore, an intelligent VM placement algorithm is needed which can adapt to the current status of PMs so that resources wastage can be minimized which alsoresults in less number of active PMs.
In this work, we propose a new VM placement algorithm for cloud computing environment. The objective of the proposed algorithms is to balance the resource utilization of active PMs and also their minimization. We adopt harmonic mean of resource utilization of VMs through which VMs with similar resources required to run are grouped together. We devise a new resource wastage model through which current resource utilization of the PMs can be measured. Finally, we simulate the proposed algorithms and simulation results are compared with the existing algorithm such as VMPACS [14]. Through performance evaluation, we establish the superiority of the proposed algorithm over the exiting algorithm.
Theremaining part of the work isorganizedasfollows.The problem formulation and the proposed algorithms are presented in Section 2. We provide a detailed descriptionandexperimental resultin Section 3. Finally, we give concluding remarks of the paper, in Section 4.
2. Proposed Work
In this section, we first define some of the terminologies used in the proposed algorithm and description of problem formulation followed by the proposed algorithms. They are presented in the following subsections consequently.
2.1 Terminologies and Problem formulation
Notations used in the problem formulation and the proposed algorithmaredescribed in Table 1. Next, we details resource wastage and power consumption models adopted by the proposed.
Total power consumption: we measure the total power consumption of the active PMs using Eq. (3), by the varying VMs from 50, 100, 150, 200, and 250. From Fig. 1, we observe that the proposed algorithm consumes 18% less power as compared to the existing algorithm. This is mainly due to the less number of PMs are active in case of proposed algorithm as compared to the exiting one, as depicted in Fig. 2. Furthermore, the results are also better due to better utilization of CPU in case of proposed algorithm as compared to VMPACS,
Total Resource wastage: we test the total resource waste of the proposed as well as exiting algorithm. In the test, we varied the number VMs form 50, 100, 150, 200 and 250. The results is shown in Fig. 4. From the figure, it is clear that, the propose algorithm minimizes resource wastage as compared to the existing algorithm. From the figure, we also observe that, the proposed algorithm minimizes the resource wastage from 47% - 82 % as compared to existing one.This is mainly due to better utilization of CPU and memory resources in the active PMs,
Abstract: Cloud computing emerged as a new computing paradigm for on-demand provisioning of IT services over the Internet. One challenging research issues in cloud environment it to develop any energy efficient and resource-aware algorithm for virtual machine placement (VMP) problem. If the problem is not dealt properly, it effects the performance of cloud data centers in terms of resource utilization and power consumption. In this paper, we propose a new VM placement for cloud environment, to address VMP problem. The objective of the proposed algorithm is maximize resource utilization and minimize the number of active physical machines. We adopt a clever strategy to assign ranks to VMs and placing VMs based on their ranks. We simulate the proposed algorithm as well as exiting algorithm. The simulation results show that the performance of the proposed is better than the existing algorithm.
1. Introduction
Cloud computing has emerged as a new computing paradigm which enables to provide computing resources to the cloud users over the Internet [1-3]. The computing resources are pooled at one place called datacenter and these centers are managed by an organization called cloud service providers. The service providers offer three types of services to the users such asInfrastructure as a Service (IaaS), Software as a Service, and Platform as a Service [4-6]. In IaaS, computing resources such as CPU, RAM, Storage and Network are offered as a service and their use is expected to abide by service level agreements [4]. Resources are shared among the users through virtualization. Virtual machines (VMs) are created over the physical machines (PMs), to execute the user's applications. One of the challenging problems in virtualization is to utilize the computing resources efficiently and minimize the power consumption. The above problem can be solved by employing an efficient virtual machine placement algorithm. The job of the placement algorithm is to place the pool of VMs on the PMs in such way that computing resources are efficiently utilized and also minimize the number of active PMs.
Numerous algorithms for VM placement have been proposed in the recent years [9-15]. However, most of the algorithms ignore the balanced utilization of resources which leads to higher resource wastage and also increase the number of active PMs. For example, PMs with d-dimensional resources, if utilization of all the resources are not balanced then it may happens that a newly arrived VM cannot be accommodated in any of the active single PM.As a consequence, increase in resource wastage of the existing active PMs and also a new PM needs to be activated in order to accommodate the newly arrived VM.This is due to the lack of an intelligent decision making while selecting VMs for placement. To demonstrate this, consider an example, as shown in Fig. 1.Virtual machines, VM1 and VM2 are placed in PM1, VM3 cannot be accommodated in PM1 and it is placed in PM2. VM4 neither housed in PM1 nor in PM2 and so it is placed in PM4. Due to improper consolidation of VMs, overall resources of the PMs are unused and also it leads to increase in number of active VMs. whereas, as shown in Fig. 2, due to proper consolidation of VMs, neither resources are wasted nor increase in number of active PMs. Therefore, an intelligent VM placement algorithm is needed which can adapt to the current status of PMs so that resources wastage can be minimized which alsoresults in less number of active PMs.
In this work, we propose a new VM placement algorithm for cloud computing environment. The objective of the proposed algorithms is to balance the resource utilization of active PMs and also their minimization. We adopt harmonic mean of resource utilization of VMs through which VMs with similar resources required to run are grouped together. We devise a new resource wastage model through which current resource utilization of the PMs can be measured. Finally, we simulate the proposed algorithms and simulation results are compared with the existing algorithm such as VMPACS [14]. Through performance evaluation, we establish the superiority of the proposed algorithm over the exiting algorithm.
Theremaining part of the work isorganizedasfollows.The problem formulation and the proposed algorithms are presented in Section 2. We provide a detailed descriptionandexperimental resultin Section 3. Finally, we give concluding remarks of the paper, in Section 4.
2. Proposed Work
In this section, we first define some of the terminologies used in the proposed algorithm and description of problem formulation followed by the proposed algorithms. They are presented in the following subsections consequently.
2.1 Terminologies and Problem formulation
Notations used in the problem formulation and the proposed algorithmaredescribed in Table 1. Next, we details resource wastage and power consumption models adopted by the proposed.
Total power consumption: we measure the total power consumption of the active PMs using Eq. (3), by the varying VMs from 50, 100, 150, 200, and 250. From Fig. 1, we observe that the proposed algorithm consumes 18% less power as compared to the existing algorithm. This is mainly due to the less number of PMs are active in case of proposed algorithm as compared to the exiting one, as depicted in Fig. 2. Furthermore, the results are also better due to better utilization of CPU in case of proposed algorithm as compared to VMPACS,
Total Resource wastage: we test the total resource waste of the proposed as well as exiting algorithm. In the test, we varied the number VMs form 50, 100, 150, 200 and 250. The results is shown in Fig. 4. From the figure, it is clear that, the propose algorithm minimizes resource wastage as compared to the existing algorithm. From the figure, we also observe that, the proposed algorithm minimizes the resource wastage from 47% - 82 % as compared to existing one.This is mainly due to better utilization of CPU and memory resources in the active PMs,