No abstract

No abstract

The Deep Learning (DL) paradigm gained remarkable popularity in recent years. DL models are used to tackle increasingly complex problems, making the training process require considerable computational power. The parallel computing capabilities o ered by modern GPUs partially fulfill this need, but the high costs related to GPU as a Service solutions in the cloud call for e cient capacity planning and job scheduling algorithms to reduce operational costs via resource sharing. In this work, we jointly address the online capacity planning and job scheduling problems from the perspective of cloud end-users. We present a Mixed Integer Linear Programming (MILP) formulation, and a path relinking-based method aiming at optimizing operational costs by (i) rightsizing Virtual Machine (VM) capacity at each node, (ii) partitioning the set of GPUs among multiple concurrent jobs on the same VM, and (iii) determining a due-date-aware job schedule. An extensive experimental campaign attests the e ectiveness of the proposed approach in practical scenarios: costs savings up to 97% are attained compared with first-principle methods based on, e.g., Earliest Deadline First, cost reductions up to 20% are obtained with respect to a previously proposed Hierarchical Method and up to 95% against a dynamic programming-based method from the literature. Scalability analyses show that systems with up to 100 nodes and 450 concurrent jobs can be managed in less than 7 seconds. The validation in a prototype cloud environment shows a deviation below 5% between real and predicted costs.

No abstract

Integrating Distributed Energy Resources (DERs) and Micro-Grid (MG) into a system evolved the traditional power system. In spite of their significant advantages, MGs may result in volatility and uncertainty in the power systems. For reliable operation of the grid, energy trading among MGs should be optimized to maintain a fair trading price, maximize participants' profit, and satisfy network constraints. In this paper, the optimal power trading among multiple reconfigurable MGs is formulated as a Mixed-Integer Nonlinear Programming (MINLP) considering all energy resources and their dynamic prices. In spite of the other methods in the literature, the proposed method minimizes the total cost (increase sales and decrease purchases) and transmission loss considering all energy resources in the MGs. In order to flatten the load profile, a time-based load profile is considered for the demand response program. The performance of the proposed model is evaluated on an IEEE 6-bus network as well as a modified IEEE 33-bus test system. The results verify that the proposed method, (i) determines the best configuration among MGs with a switching reduction of about 30%, (ii) optimizes the power generation of energy resources with 12% reduction in energy production, and (iii) optimizes the power trading costs with a 10% reduction in costs compared with the basic model without DR and trade that is introduced as Scen.1 in this paper.