Academia.eduAcademia.edu

Figure 3 – uploaded by Utpal Biswas

See full PDFdownloadDownload figure
Figure 6 Before increment of “‘Balance” value. In Fig. 6, at first assume the initial value of Balance is set to 2. After allocating cloudlets CO to C5 to the VMs the number of cloudlets allocated to each VM becomes equal to the current Balance value but still new cloudlet C6 arrives. Then ADCB

Figure 6 Before increment of “‘Balance” value. In Fig. 6, at first assume the initial value of Balance is set to 2. After allocating cloudlets CO to C5 to the VMs the number of cloudlets allocated to each VM becomes equal to the current Balance value but still new cloudlet C6 arrives. Then ADCB

Related Figures (16)

It allocates the cloudlet to first available VM. For example, con- sider there are four cloudlets (Co, C1, Co, C3, and C4) and three
It allocates the cloudlet to first available VM. For example, con- sider there are four cloudlets (Co, C1, Co, C3, and C4) and three
Table 2. A comparative study of existing policies. Several Grid simulators (George Amalarethinam and Muthulakshmi (2011)) such as GridSim, SimGrid, and Gang- Sim are used to simulate the Grid application in a distributed environment, but don’t support the infrastructure and application-level requirements to simulate Cloud computing paradigm (Khanli and Analoui, 2008). A Cloud computing environment simulator must support for real-time trading of services between customers and providers. The CloudSim framework (Belalem et al., 2010) which is an open source cloud computing environment simulator is developed on GridSim toolkit (Bhatia et al., 2010). The CloudSim supports resource management and application scheduling simulation and imple- mentation of new policies. The CloudSim provides a series of extended classes and methods. It also helps to analyze new cloudlet allocation policies and scheduling criteria at different levels. The CloudSim supports modification of in built classes, module deployment techniques and performance analysis by Min-—min algorithm lacks uniform resource utilization in a way that it chooses smaller cloudlets first which makes use of VMs with higher processing speed. Therefore, the scheduling is not optimal when the number of smaller cloudlets is greater than the large ones. To overcome this drawback (Armbrust M et al., 2010), Max—min algorithm schedules larger cloudlets at first. But sometimes, due to the prior execution of larger cloudlets the makespan may increase. Max—min also increases the waiting time of smaller cloudlets. RASA has disadvantages of both Min—min and Max-—min, although it approaches to reduce them. In RRA, large cloudlets are often assigned to the VMs with low MIPS and hence take a longer time to exe- cute as well as increasing the waiting time and the response
Table 2. A comparative study of existing policies. Several Grid simulators (George Amalarethinam and Muthulakshmi (2011)) such as GridSim, SimGrid, and Gang- Sim are used to simulate the Grid application in a distributed environment, but don’t support the infrastructure and application-level requirements to simulate Cloud computing paradigm (Khanli and Analoui, 2008). A Cloud computing environment simulator must support for real-time trading of services between customers and providers. The CloudSim framework (Belalem et al., 2010) which is an open source cloud computing environment simulator is developed on GridSim toolkit (Bhatia et al., 2010). The CloudSim supports resource management and application scheduling simulation and imple- mentation of new policies. The CloudSim provides a series of extended classes and methods. It also helps to analyze new cloudlet allocation policies and scheduling criteria at different levels. The CloudSim supports modification of in built classes, module deployment techniques and performance analysis by Min-—min algorithm lacks uniform resource utilization in a way that it chooses smaller cloudlets first which makes use of VMs with higher processing speed. Therefore, the scheduling is not optimal when the number of smaller cloudlets is greater than the large ones. To overcome this drawback (Armbrust M et al., 2010), Max—min algorithm schedules larger cloudlets at first. But sometimes, due to the prior execution of larger cloudlets the makespan may increase. Max—min also increases the waiting time of smaller cloudlets. RASA has disadvantages of both Min—min and Max-—min, although it approaches to reduce them. In RRA, large cloudlets are often assigned to the VMs with low MIPS and hence take a longer time to exe- cute as well as increasing the waiting time and the response
Figure 5 Cloudlets in CSNA phase. It is the last phase of RB2B. After the first two phases if the bal- ancing factors decide to allocate the arriving cloudlet to a VM, ADCB will move on to the next cloudlet. But if the arriving
Figure 5 Cloudlets in CSNA phase. It is the last phase of RB2B. After the first two phases if the bal- ancing factors decide to allocate the arriving cloudlet to a VM, ADCB will move on to the next cloudlet. But if the arriving
Table 3. Cloudlet length acceptability ranges for each VM. and Cmax are the minimum and maximum cloudlet length (In MI). MIPS; is the processing speed of VM;. So, the total of the processing speeds of the VMs is
Table 3. Cloudlet length acceptability ranges for each VM. and Cmax are the minimum and maximum cloudlet length (In MI). MIPS; is the processing speed of VM;. So, the total of the processing speeds of the VMs is
Figure 7 After increment of “Balance” value.
Figure 7 After increment of “Balance” value.
Figure 8 Variation of local queue length of VMs.
Figure 8 Variation of local queue length of VMs.
Table 7 RB2B allocation procedure.
Table 7 RB2B allocation procedure.
Table 5 Reference VM.
Table 5 Reference VM.
This section deals with analyzing the improvement of the results of the RB2B compared to RASA, Max-—min, Min— min, RRA and CA. A reference stream of 1000 cloudlets with cloudlet length varying from 1000 MI to 100,000 MI and 10 VMs with processing speed varying from 1000 MIPS to 10,000 MIPS is used for the comparison. Due to space con- straints, only this experimental scenario is being considered, but in experiments with larger number of cloudlets and VMs identical results have been found. The simulated results are evaluated and analyzed in several aspects. The performance was measured by setting up simulation environments (36) in Figure 12 Average make span comparison.
This section deals with analyzing the improvement of the results of the RB2B compared to RASA, Max-—min, Min— min, RRA and CA. A reference stream of 1000 cloudlets with cloudlet length varying from 1000 MI to 100,000 MI and 10 VMs with processing speed varying from 1000 MIPS to 10,000 MIPS is used for the comparison. Due to space con- straints, only this experimental scenario is being considered, but in experiments with larger number of cloudlets and VMs identical results have been found. The simulated results are evaluated and analyzed in several aspects. The performance was measured by setting up simulation environments (36) in Figure 12 Average make span comparison.
Table 9 Percentage of Improvement.
Table 9 Percentage of Improvement.
Table 8 Comparison result for performance evaluation.
Table 8 Comparison result for performance evaluation.
CloudSim 3.0.3. The improvement of completion time and makespan is explained with an interpretation of the graphs in tabular form. Due to space constraint, the details of the reference cloudlet string and the VMs could not be given here. The improvement in the result of the proposed work indicates the efficiency of RB2B. Five parameters are taken into account to compare and analyze the performance of these algorithms. They are average waiting time (AWT), average turnaround
CloudSim 3.0.3. The improvement of completion time and makespan is explained with an interpretation of the graphs in tabular form. Due to space constraint, the details of the reference cloudlet string and the VMs could not be given here. The improvement in the result of the proposed work indicates the efficiency of RB2B. Five parameters are taken into account to compare and analyze the performance of these algorithms. They are average waiting time (AWT), average turnaround
Table 10 A comparative study of existing algorithms & proposed RB2B algorithm.
Table 10 A comparative study of existing algorithms & proposed RB2B algorithm.
Related topics:
Computer Science
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved