Model for VM AllocationRecent advancement of cloud usage increased rapidly. Resource provisioning and allocation policies updatedperiodically when increasing number of cloud users. Efficient use of cloud resources, virtualization technology getsinvolved and support for multi-users. This is achieved by dynamic placement of VMs and live migration scheme.Dynamic VM placement is most powerful technique used to allocate required VMs on physical machines. Hence, energyefficient VMs placement increases the performance of the host; meanwhile unused host goes to sleep state to reduce energyconsumption.3.3.1 VM SelectionAllocation of VM is divided into two major categories such that, VM selection and VM placement. First part we haveto select over utilization and under utilization VM from each host using SLA-Aware MBFD algorithm. Second part,selected VM has to be migrated using EMPMU policy. To find total utilization of CPU and power consumption, we setlower and upper threshold value to get efficient results. When threshold values exceeds from the allocated limit, it willconcludes VM migration is required for those host. If the threshold value goes to lower limit, we consider as VMs migratedto idle state or sleep mode to minimize power consumption. Sometimes, over utilized VMs migrated to underutilized hostto balance the server load.3.3.2 VM PlacementConsolidation of VMs has more close to bin packing problem, where each VM has varying size (bin size) and resource(items) requirement. In our proposed scheme, we focus on efficient utilization of resources and consolidating VMs usingSLA-Aware MBFD algorithm. The basic terminology of bin packing problem introduced by Martello et al. 22 and it canbe described as follows,Given n items (VM) and m bins (host). Let as consider set of host denoted by H1, H2, H3,c…….,.Hk with same size C.List of n VMs and weight should be consider as VM1,VM2,VM3,c…………,VMn. Assign each VM to one host so that totalweight of the VM in each host does not exceed C and the number of host used is minimum. Find an integer number of hostB and B-partition considered as H1UH2c..UHk such that,(11)Where k=1,2,……..,B. Solution of B is optimal when it is used minimum hosts. The B value of an optimal solutionconsidered as OPT. A possible mathematical formulation of the problem considered as 23:(12)Subject to B.1,(13)(14)Where yi = 1, if host i is used and xij = 1 if VM j is placed in host i. The weights of VMj are positive integers. Therefore,without loss of generality, we can define C is a positive integer.3.3.3 Enhanced minPower and maxUtilization(EMPMU) VM Migration PolicyThe optimization of VM migration is carried out using upper and lower bound of CPU utilization. Migration of eachVM has been calculated by total resource usage of current CPU utilization. To migrate VMs from one host to another hostwe have arise these following questions: when, where and which VM to migrate. Here, we have used double-threshold VMmigration policy to reduce power consumption. Migration of VMs addressed with two conditions. First, when CPUutilization goes to reach lower threshold, all VMs moved from current host to sleep state. Meanwhile, CPU utilizationexceeds upper bound, those VMs moved to minimum utilization host and applying consolidation techniques. Henceforth,the objective of VM migration is reduces energy consumption and utilizes heterogeneity resource effectively.Algorithm 1. Enhanced minPower and maxUtilization(EMPMU) Algorithm1 Input: VMsToMigrate, minPower2 Output: MigrationMap3 AscendingVMs=sort_Ascending(VMsToMigrate)4 DecendingVMs=sort_Decending(VMsToMigrate)5 foreach Vj in DecendingVMs do6 AllotedHost=minPower(Vj)7 AscendingVMs.remove(Vj)8 DecendingVMs.remove(Vj)9 foreach Vk in AscendingVMs do10 if AllotedHost.isHostUnder_Utilization(Vk) then11 waitState.put(allotedHost, Vk), goto step1312 else if AllotedHost.isHostOver_Utilization(Vk) then13 MigrationMap.put(allotedHost,Vk)14 allotedHost.create(Vk)15 AscendingVMs.remove(Vk)16 DecendingVMs.remove(Vk)17 else18 end19 return MigrationMap3.4 Model for Resource Allocation AlgorithmAn increasing number of cloud users expecting uninterrupted services over internet. Those, services provided by theCloud Service Providers (CSP) and they are mainly focuses on increasing resources utilization and Return on Investment(ROI). In this scenario, we have proposed efficient VM scheduling algorithm to manage all those resources in cloud datacenters. Whenever, new jobs arrived into the VM, VM manager monitor the current workload of VM. Then, decidesallocate the jobs on the same VM or create new one based on the user requirements.3.4.1 SLA-Aware Resource AllocationThe problem of energy efficient VM provision and allocation is always challenging part for cloud providers. Whilemeeting SLA should emphasized in each level and providers cannot change their convenient consolidation or migrationprocess. For privileged users, providers should satisfy QoS without violating SLA. Our proposed scheme defined VMprovision using SLA-Aware MBFD algorithm. Moreover, place the VM in an appropriate host using resource usage basedEMPMU migration technique.3.4.2 SLA-Aware Modified Best Fit Decreasing(SLA-Aware MBFD) AlgorithmReducing computational complexity and increasing optimal solution, many heuristic algorithms developed like, first-fit,best-fit and worst-fit. Each algorithm produces different non-guarantee optimal solution depends on the number of objectsplaced in each bin. Hence, effective use of heuristic algorithm for VM placement, we applied SLA-Aware MBFDalgorithm. The proposed model we have applied 11/9.OPT+1 where OPT is the number of bins (Host) given by the optimalsolution 23. When allocate VM in each host it will take least unit of power in an increasing order. Due to this allocationmodel, VM consume more energy and increasing operational cost of the data centers. In MFBD algorithm sort, all the VMsin decreasing order based on current CPU utilization of a machine. Furthermore, it will allocate power efficient nodes first.Apart from traditional MFBD, we have enhanced our algorithm with SLA constraints represent in Table 1. Here, lowerlevel SLA assigned to minimum resource usage and higher-level assigned to maximum resource requirement. Let n isconsider as total number of SLA levels and L is an individual level. SLA allocation constraints addressed from 0 to n-1 is0.SLA_Level.n-1. Without compromising SLA violations, our scheduling algorithm works efficiently and it givesminimum energy consumption and maximum resource utilization.Algorithm 2. SLA-Aware Modified Best Fit Decreasing(SLA-Aware MBFD) Algorithm1 Input: hostList, vmList, SLA_List,n2 Output: allocation of VMs3 vmList.sortDecreasingUtilization()4 foreach vm in vmList do5 minPower©MAX6 allocatedHost©NULL7 foreach host in hostList do8 if host has enough resource for vm then9 foreach SLA_Level in SLA_List do10 if SLA_Level.n-1 then11 SLA_Lower©lowerCPU_Threshold12 SLA_Higher©higherCPU_Threshold13 power©estimatePower(host, vm)14 if power < minPower then15 allocatedHost©host16 minPower©power17 if allocatedHost ‚ NULL then18 allocate vm to allocatedHost19 return allocation4. Performance Analysis and Results DiscussionEfficient resource allocation heuristics using SLA-Aware MBFD and Enhanced minPower, maxUtilization migrationpolicy presented in section 3. Our experimental results increase resource utilization with less SLA violation and reducepower consumption in cloud data centers. To validate our proposed model, we have chosen CloudSim 3.0.3 simulationtoolkit. Moreover, the reason for choosing CloudSim toolkit supports multi-objective functions in single platform like,resource allocation, VM placement, power management, network, and storage management.