Two-Phase Virtual Machine Placement in Cloud Computing Data Centers Using Extreme Learning Machine Prediction-Based Triggering Method

Nafiu Musa Muhaammad, Faruk Umar Ambursa and Nasir Faruk

Keywords:Virtual machine placement, Cloud computing, Incremental VMP, VMP reconfiguration


Background: Two-phase Optimization of Virtual Machine Placement (VMP) Problem considers both the Online
Incremental VMP (iVMP) phase in which the new arrival of dynamic requests of Virtual Machines VMs are attended to
and the Offline VMP reconfiguration (VMPr) phase that performs placement recalculation. In the two-phase scheme, the
first part of the two-phase approach is the iVMP, where virtual machines (VMs) can be built, changed, or destroyed at
runtime. While the second phase focuses on raising the standard of solutions produced by the iVMP, several studies have
been done in different literature to solve the VMP problem. However, the methods used tend to be over-forecast and have
long runs of a linear trend. This affects the prediction and produces a less optimal solution. Objective: The following four
objective functions are optimized using the proposed Extreme Learning Machine Prediction-Based Triggering Method
for Virtual Machine Placement in Cloud Computing Datacenters in Two-Phases, which combines the advantages of both
online (dynamic) and static (offline) VMP formulations: the length of the reconfiguration process, the amount of energy
used, the way resources are used, and the financial expenses. This study suggests a brand-new strategy for deciding when
to start the VMP reconfiguration phase. Results: The Method provides more accuracy to the predicted requests as well
as reduces the total economic penalties for Service Level Agreement (SLA) violations. An experimental comparison with
the existing approach is conducted utilizing 400 cases. Conclusion: The results demonstrated that, in comparison to the
benchmark approach, the proposed work obtained a minimum cost function with a 10.5% improvement.