1,720,965 research outputs found
Power efficient resource allocation in cloud computing data centers using multi-objective genetic algorithms and simulated annealing
No full textIn this paper we present a new task allocator for Cloud Data Center (DC). The implementation is based on two different heuristics: Multi-Objective Genetic Algorithms (Moga) and Simulated Annealing (SA). The allocator reduces at the same time both task completion time and server and switches power consumption, avoiding network link congestion. The evaluation results show that the developed approach is able to perform the static allocation of a large number of independent tasks on homogeneous single-core servers with a quadratic time complexity for Moga and a linear time complexity for SA
Multi Objective Virtual Machine Allocation in Cloud Data Centers
No full textIn this paper, we propose a Virtual Machine (VM) allocator for Cloud Computing Data Center (DC). We allocate a set of VMs on servers that are interconnected through a three-tier fat-tree network topology. VMs require four different resources: CPU, memory, disk, and bi-directional network bandwidth forcommunications directed to and coming from the external gateway. Our goal is not to overload computing devices (i.e. allocating more resource than servers' availability) while reducing servers and switches power consumption, in the current proposal, power consumption of each device follows a load-proportional trend. The allocation problem is combinatorial and non-convex, and it is a variant of the multi objective bin packing problem which is NP-Hard. For these reasons, we solve the problem using a particular kind of heuristics called Multi Objective Genetic Algorithm (MOGA) and inspired by the natural process of evolution, MOGA is quite often able to effectively approximate complex problems, such us the one considered. We perform a comparison with a simplified and single-objective formulation of the problem that is solved using CPLEX, while solutions are evaluated using specific quality indicators. The results show how the presented approach solves the allocation problem: MOGA retrieves good quality solutions in less than ten seconds allocating thousands of Vms and obtaining the sameresults as CPLEX
Shooter localization in wireless acoustic sensor networks: A feasibility study of a low cost and power effective wireless acoustic sensor network for shooter localization
No full textThis paper introduces a feasibility study of a low cost and power effective wireless acoustic sensor network for shooter localization. Currently deployed sensors for this application use a time domain gunshot signal analysis and high sampling rates, in the order of MS/s. We investigate an alternative Short Time Fourier Transform with a lower sampling rate (250 kS/s). We focus on a single channel single sensor approach for shooter ranging and we provide new experimental data. We implement a centralized gateway on top of the open and programmable board named ZedBoard. Our approach localizes the shooter with a distance error that ranges between roughly half and three meters, so the results are encouraging for further studies
A novel SDN controller for traffic recovery and load balancing in data centers
No full textNowadays, Software-Defined Networking (SDN) is a powerful architectural model, decoupling control and forwarding planes through the abstraction of network elements and functionalities. As known, in SDN networks the controller is the key element since the intelligence of the network is centralized. Hence, the deployment of network services, such as QoS, traffic engineering and traffic recovery, requires the design and development of control apps. In this paper we focus on load balancing and traffic recovery, specifying a controller architecture and integrating the above-mentioned services on top of the POX platform. Moreover, different levels of protection are implemented depending on the QoS guarantees required by the flow (on per class-of-service basis). To validate the behaviour of our controller, we considered a network with fat-tree topology, widely deployed in modern data centers: after the functional assessment, load balancing performance have been compared for different traffic flows and cost assignment strategies
A Power Efficient Genetic Algorithm for Resource Allocation in Cloud Computing Data Centers
Full text linkpeer reviewe
Throughput Maximization Scheduling Algorithm in TSCH Networks with Deadline Constraints
No full textTime Slotted Channel Hopping (TSCH), defined among the operating modes in IEEE 802.15.4-2015 standard, was established to offer a guaranteed quality of service for deterministic industrial type applications. However, the standard only provides a framework but it does not mandate a specific scheduling mechanism. In this paper, we formulate the NP-hard scheduling problem in terms of maximizing the throughput with deadline constraints and at the same time satisfying interference constraints in TSCH Networks. In the considered TSCH network, a centralized entity typically called gateway, coordinates the assignment of frequencies and timeslots to the nodes. To solve this NP-hard throughput scheduling problem, a Genetic Algorithm (GA) framework was adopted. Simulation results corroborate that our GA-based approach yields very close performance to the optimal solutions and operates with much lower complexity. In addition, the results also confirmed that GA outperforms other popular scheduling algorithms in the literature in terms of throughput maximization as well minimizing violated deadlines
A novel allocation strategy for virtual machines in software defined data center
No full textCloud Data Center (DC) service orchestration and resource management are important applications of the Software Defined Networking (SDN) paradigm. In this paper, we introduce a novel dynamic allocation strategy for Virtual Machine (VM) allocation called Enhanced multi-objective Worst Fit (E-WF). E-WF combines the multi-objective Best Fit and Worst Fit allocation strategies, and it exploits the history of the previous requests to limit the resource fragmentation. We allocate the VMs choosing jointly a server and the least power-consuming network path. We show in our simulations that E-WF performs better and allocates more VMs with respect to the other presented approaches, and the power-aware network path allocation reduces the power consumption of network devices with respect to the classical shortest path first routing algorithm
A fuzzy logic approach for resources allocation in cloud data center
No full textThe rapidly increasing demand of Cloud services, asking for a flexible and dynamic design of the Cloud, has become a major challenge in DC deployment. Classical Traffic Engineering approaches are no longer enough to deal with the efficient use of IT and network resources in this highly distributed scenario. To cope with this issue, we propose two Fuzzy logic controllers for DC resource allocation based on Mamdani and Sugeno inference processes, that are able to take advantage of simple heuristic rules for efficient virtual machines allocation. To test the effectiveness of the proposed controllers we compare their performance with two variants of Multi-objective allocators as well as a simple Mono-dimensional algorithm. Preliminary simulation tests validate our proposal in terms of number of allocated requests and average server resource utilization
Power Consumption-Aware Virtual Machine Placement in Cloud Data Center
Full text linkIn this paper, we present a set of power-aware dynamic allocators for virtual machines (VMs) in cloud data centers (DCs) taking advantage of the software defined networking paradigm. Each VM request is characterized by four parameters: 1) CPU; 2) RAM; 3) disk; and 4) bandwidth. We design the allocators in order to accept as many VM requests as possible, taking into account the power consumption of the network devices. In this paper, we introduce ten different allocation strategies, and compare them with a baseline that consists of using the first available server (first fit). The allocators differ in terms of allocation policy (best fit/worst fit), allocation strategy (single/multi objective optimization), and joint/disjoint selection of IT and network resources. For both joint and disjoint approaches, we evaluate the behavior of all possible pairs policy-strategy, varying the load of the DC and the number of VMs to be allocated. Moreover, the experimental results highlight that joint approaches outperform disjoint ones
Power consumption-aware virtual machine allocation in cloud data center
No full textThis paper compares a set of Virtual Machine (VM) allocators for Cloud Data Centers (DCs) that perform the joint allocation of computing and network resources. VM requests are defined in terms of system (CPU, RAM and Disk) and network (Bandwidth) resources. As concerns the first ones, we allocate VM resources following two different policies, namely Best Fit and Worst Fit, corresponding to consolidation and spreading strategies respectively. For each server, the allocators choose the network path that minimizes electrical power consumption, evaluated according to a precise model, specifically designed for network switches. More specifically, we implemented different allocation algorithms based on Fuzzy Logic, Single and Multi-Objective optimization. Simulation tests have been carried out to evaluate the performance of the allocators in terms of number of allocated VMs for each policy. Finally, it is worth mentioning that we have designed the proposed allocators as a building block of a Software Defined Networking (SDN) orchestrator
- …
