Therefore, the feature model presented in this study can use this estimated time as the input data to perform a static task placement and scheduling in TPVM. We further discuss the execution of TPVM and perform the experiments in CloudSim in next section.
TPVM algorithm is developed in the source of CloudSim to extend the function of platform, thus, to verify the properties of TPVM.
In the implementations of TPVM, three new classes Cloudet_unit, Cloudet_thread and CloudetFMextended from Cloudet is designed, and the tuple of Unit, Cloudet_unit ,and the tuple of FM are added into these classes as the members of themselves.
We design a new class DatacenterBroker_FourLayers extended from DatacenterBroker, whose function bindCloudletToFourLayers(Cloudet_FM fm, Parameter para) implements TPVM algorithm to solve the problem in the model of task placement and scheduling based on virtual machines.
Two steps in below section were made to verify the rationality of TPVM. Our experiments aim to determine the effect that physical configuration and personal preference have on the convergence of TPVM.
When the input parameters of TPVM algorithm are fixed, (eg.
We can see from the result that TPVM could find out the optimal solution for any group of physical configuration and the optimal solution is in the lowest point of any of such four Time curves or the highest point of any of such four MD curves, which are consistent with the iterative nature of TPVM.
We can see from the result that TPVM could find out the optimal solution for any group of personal preference and the optimal solution is in the lowest point of any of such four Time curves or in the highest point of any of such four MD curves, which are consistent with the iterative nature of TPVM.
It is concluded from above experiment that TPVM algorithm can find out a group of [pi], [xi], [zeta] to promote the value of Time come to their minimum value and MD come to their maximum value for any personal preference and physical configuration, so the TPVM algorithm is feasible to implement task placement and scheduling.
It is shown that TPVM can improve the matching degree between tasks and architectures to raise the peak performances of clusters and satisfy the requirement of designers in applications development and deployment.
The implementation of TSA is as follows: On the basis of designing the new classes Cloudet_unit, Cloudet_thread,Cloudet_FM and DatacenterBroker_FourLayers, Stastics in above section, according to results of TPVM, an new function executeCloudletInFourLayers (Cloudet_FM fm, Threads, VirtualMachineList VMs, host PMs) is added into the class DatacenterBroker_FourLayers to implement the TEVM.
In TPVM algorithm, the time estimated values in FM has been inputted into TPVM.