Resource Management In Data Centers Research Articles

Energy Aware Resource Management of Cloud Data Centers

Cloud Computing, the long-held dream of computing as a utility, has the potential to transform a large part of the IT industry, making software even more attractive as a service and shaping the way IT hardware is designed and purchased. Virtualization technology forms a key concept for new cloud computing architectures. The data centers are used to provide cloud services burdening a significant cost due to high energy consumption. Data centers are provisioned to accommodate peak demand rather than average demand and cloud applications consume much more electrical energy than they need. Thus, it necessitates that cloud computing solutions not only minimize operational costs, but also reduce the power consumption. In this paper, we investigate load balancing and power saving methods in virtualized cloud infrastructures. Imbalanced distribution of workloads across resources can lead to performance degradation and much electrical power consumption in such data centers. We present an architectural framework and principles for energy-efficient cloud computing environments. Resource provisioning and allocation algorithms, named Load-Power-aware, are proposed in this architecture. The algorithm employs a heuristic to dynamically improve the energy efficiency in data center, while guarantees the Quality of Service (QoS). The efficiency of the proposed approach is evaluated by using the most common cloud computing simulation toolkit, CloudSim. The performance modeling and simulation results are depicted the proposed approach significantly improves the energy efficiency in a given dynamic scenario, while a small amount of service level agreements (SLA) is missed.

Rate-based thermal, power, and co-location aware resource management for heterogeneous data centers

Today’s data centers contain large numbers of compute nodes that require substantial power, and therefore require a large amount of cooling resources to operate at a reliable temperature. The high power consumption of the computing and cooling systems produces extraordinary electricity costs, requiring some data center operators to be constrained by a specified electricity budget. In addition, the processors within these systems contain a large number of cores with shared resources (e.g., last-level cache), heavily affecting the performance of tasks that are co-located on cores and contend for these resources. This problem is only exacerbated as processors move to the many-core realm. These issues lead to interesting performance-power tradeoffs; by considering resource management in a holistic fashion, the performance of the computing system can be maximized while satisfying power and temperature constraints. In this work, the performance of the system is quantified as the total reward earned from completing tasks by their individual deadlines. By designing three resource allocation techniques, we perform a rigorous analysis on thermal, power, and co-location aware resource management using two different facility configurations, three different workload environments, and a sensitivity analysis of the power and thermal constraints.

Fast multi-resource allocation with patterns in large scale cloud data center

How to achieve fast and efficient resource allocation is an important optimization problem of resource management in cloud data center. On one hand, in order to ensure the user experience of resource requesting, the system has to achieve fast resource allocation to timely process resource requests; on the other hand, in order to ensure the efficiency of resource allocation, how to allocate multi-dimensional resource requests to servers needs to be optimized, such that server's resource utilization can be improved. However, most of existing approaches focus on finding out the mapping of each specific resource request to each specific server. This makes the complexity of resource allocation problem increases with the size of data center. Thus, these approaches cannot achieve fast and efficient resource allocation for large-scale data center. To address this problem, we propose a pattern based resource allocation mechanism based on the following findings. In a real-world cloud environment, the resource requests are usually classified into limited types. Thus, the mechanism first utilizes this feature to generate pattern information, which indicates which types of resource requests are suitable to be allocated together to a server. Then, the mechanism uses the pattern information as guidelines to make fast resource allocation decision and fully utilize server's multidimensional resources. Simulation experiments based on real and synthetic traces have shown that our mechanism significantly improves system's resource utilization and reduces the overall number of used servers.

Pre-Filter-Copy: Efficient and Self-Adaptive Live Migration of Virtual Machines

Live migration of virtual machines (VMs) is useful for resource management of data centers and cloud platforms. The precopy algorithm is widely used for memory migration. However, when encountered with write-intensive workloads, the precopy's straightforward iteration strategy will become inefficient. Worse still, it is hard to tune the performance with the existing parameters, unless load characteristics are known in advance. In this paper, we propose an improved pre-filter-copy (PFC) algorithm. The main target is to reduce migration time and bandwidth resource consumption of the precopy algorithm, while keeping downtime at the same level. We designed a novel data filter to achieve this goal. In each round of iteration, it forecasts the pages that will be subsequently dirtied and then filters them from the send list. Meanwhile, previously filtered pages will be reconsidered, to see if they can be added to the send list. This ensures that the downtime will not be increased. Furthermore, a new parameter is proposed to improve the adaptivity of the precopy algorithm. Experimental results show that the PFC algorithm significantly reduces migration time and the amount of migrated data, while keeping the downtime at the same level.

Novel fuzzy multi objective DVFS-aware consolidation heuristics for energy and SLA efficient resource management in cloud data centers

Cloud computing is an attractive solution for managing IT infrastructures due to its numerous beneficiaries such as automatic optimum resource management as well as modern service delivery models. Cloud computing relies on huge energy hungry data centers to supply resources for diverse range of users with dynamic workloads. Hence, reducing the ever-increasing energy consumption of Cloud data centers as well as their carbon footprints is a crucial issue for Cloud providers. This paper takes advantage of dynamic voltage and frequency scaling method as well as consolidation approach to propose a novel fuzzy multi criteria and objective resource management solution. The extensive evaluation of proposed algorithms using Cloudsim simulator shows notable reductions in energy consumption, SLA violation, and number of migrations in comparison with state of the art.

Optimized Energy Efficient Virtual Machine Placement Algorithm and Techniques for Cloud Data Centers

Cloud computing is an internet based computing technology that provide on demand computing for end users. Normally, data centers allocation for application on statically based. But today so many data centers have a problem how to reduce energy consumption? Due to increase use of cloud services and infrastructure by various cloud providers, uses of energy day by day increase that's why energy consumption increase lots. Large numbers of data centers that consume lots of energy which increase the level of co2. Hence there is need for green computing and for energy efficient management of cloud data center resources besides meeting the QoS constraints. Within a data center, maximum energy is consumed by cooling systems and ICT Infrastructure, particularly, the servers. So it is of utmost importance to optimize energy utilization in datac enter servers. In order to achieve this, we can leverage the power of virtualization which is intrinsic in Cloud Computing. Virtualization opens the doors for VM consolidation by allowing dynamic migration of virtual machines across physical machines. In this study, the problem of VM consolidation has been formulated as a mathematical optimization problem.

Penalty‐aware and cost‐efficient resource management in cloud data centers

SummaryToday, data centers are the main source of providing cloud services through a service level agreement (SLA). Most research papers for cloud resource management concentrate on how to reduce host energy consumption and SLA violation (SLAV) to minimize operational cost. However, they do not consider the amount of penalty that cloud provider should pay to users because of SLAV. In this paper, we propose a new penalty‐aware and cost‐efficient method that considers cloud resource management as a cost problem. In this method parameters such as user budget, penalty, and host energy consumption cost play an important role in minimizing operational cost which leads to higher profit for cloud provider. The simulation results with CloudSim show that our proposed method minimizes operational cost compared to the prior resource managements. Copyright © 2016 John Wiley & Sons, Ltd.

Workload-aware resource management for software-defined compute

With advance of cloud computing technologies, there have been more diverse and heterogeneous workloads running on cloud datacenters. As more and more workloads run on the datacenters, the contention for the limited shared resources may increase, which can make the management of the resources difficult, often leading to low resource utilization. For effective resource management, the management software for the datacenters should be redesigned and used in a software-defined way to dynamically allocate "right" resources to workloads based on different characteristics of workloads so that they can decrease the cost of their operation while meeting the service level objectives such as satisfying the latency requirement. However, recent datacenter resource management frameworks do not operate in such software-defined ways, thus leading to not only the waste of resources, but also the performance degradation. To address this problem, we have designed and developed a workload-aware resource management framework for software-defined compute. The framework consists mainly of the workload profiler and workload-aware schedulers. To demonstrate the effectiveness of the framework, we have prototyped the schedulers that minimize the interferences on the shared computing and memory resources. We have compared them with the existing schedulers in the OpenStack and VMWare vSphere testbeds, and evaluated its effectiveness in high contention scenarios. Our experimental study suggests that the use of our proposed approach can lead to up to 100 % improvements in throughput and up to 95 % reductions in tail latency for latency critical workloads compared to the existing ones.

Enhancing Performance of Applications in Cloud using Hybrid Scaling Technique

Infrastructure as a Service (IaaS) model of cloud computing paradigm, users acquire computing resources such as CPU, memory, storage and network bandwidth from an IaaS provider and these resources are used to deploy and run their applications. Cloud service providers share computing resources of a physical machine by running isolated Virtual Machines (VM) for web applications. As the load on web application increases, the associated VM must be able to scale up resources to support the increasing load. At the same time, VM should also be able to scale-down resources at light load. In this paper the novel architecture is proposed that provides the hybrid solution of vertical followed by horizontal scaling techniques of resource management in cloud data center. As per the dynamic load on web applications, the proposed algorithm takes the appropriate scaling decision to allocate resources from available pool of resources. Generally dynamic scaling is achieved by the conventional live VM migration technique to create additional VM instances, but VM migration spends CPU time and consumes large amount of IO and network traffic. The proposed technique postpones live VM migration as long as possible with the help of vertical scaling technique to improve the performance of applications. KeywordsComputing, Infrastructure as a Service, Virtualization, Cloud Resource Management, Virtual Machine Migration, Horizontal Scaling, Vertical Scaling

Virtual resource prediction in cloud environment: A Bayesian approach

With increase in requirement for dynamic execution of user׳s application in cloud, resource prediction techniques are gaining a lot of importance as the foundation for online capacity planning and virtualized resource management in data centers. There is a wide scope for the development of accurate resource requirement prediction methods to ensure that the virtualized resources do not suffer from over or under-utilization. We propose a Bayesian model to determine short and long-term virtual resource requirement of the CPU/memory intensive applications on the basis of workload patterns at several data centers in the cloud during several time intervals. However, the model is applied to predict resource(s) of all applications in general.The parameters considered for prediction in the model are day of week, time-interval of application access, workload, benchmarks, and availability of virtual machines etc. The model is simulated by using the SamIam Bayesian network simulator and workload traces of Amazon EC2 and Google CE data centers in dynamic scenarios. The performance is evaluated by considering benchmarks of CPU intensive applications (web based). The proposed model is able to predict virtual resources in a cloud environment with better accuracy as compared to other models.

Optimizing Resource Utilization of a Data Center

To provision IT solutions with reduced operating expenses, many businesses are moving their IT infrastructures into public data centers or are starting to build their own private data centers. Data centers can provide flexible resource provisioning in order to accommodate the workload demand. In this paper, we present a comprehensive survey of most relevant research activities on resource management of data centers that aim to optimize the resource utilization. We first describe the resource overprovisioning problem in current data centers. Then, we summarize two important components in the resource management platform and present the benefit of accurately predicting the workload in resource management. Afterwards, we classify existing resource management in a data center into three categories: 1) virtual machine-based, 2) physical machine-based, and 3) application-based resource management mechanisms. We discuss the performance degradation for implementing these three kinds of resource management in a heterogeneous data center. Finally, we present three important issues arose in the data center resource management and some potential approaches to address the issues. This paper presents a timely survey on resource management in a data center, and provides a comprehensive reference for further research in this field.

Correlation-based virtual machine migration in dynamic cloud environments

Virtual machine (VM) migration enables flexible and efficient resource management in modern data centers. Although various VM migration algorithms have been proposed to improve the utilization of physical resources in data centers, they generally focus on how to select VMs to be migrated only according to their resource requirements and ignore the relationship between the VMs and servers with respect to their varying resource usage as well as the time at which the VMs should be migrated. This may dramatically degrade the algorithm performance and increase the operating and the capital cost when the resource requirements of the VMs change dynamically over time. In this paper, we propose an integrated VM migration strategy to jointly consider and address these issues. First, we establish a service level agreement-based soft migration mechanism to significantly reduce the number of VM migrations. Then, we develop two algorithms to solve the VM and server selection issues, in which the correlation between the VMs and the servers is used to identify the appropriate VMs to be migrated and the destination servers for them. The experimental results obtained from extensive simulations show the effectiveness of the proposed algorithms compared to traditional schemes in terms of the rate of resource usage, the operating cost and the capital cost.

Automatic energy efficiency management of data center resources by load-dependent server activation and sleep modes

The operation of large Data Centers (DC) with thousands of servers is very costly in terms of energy consumption and cooling requirements. Currently, major efforts can be observed for server virtualization and consolidation to approach a proportionality between computation amount and energy consumption. In this contribution, a generalized model is presented which allows an automatic server consolidation by a load-dependent control of server activations using multi-parallel hysteresis thresholds, cold and hot server stand-by, and Dynamic Voltage and Frequency Scaling (DVFS). For the energy-efficiency and performance analysis, a multi-server queuing model is defined which is controlled by a Finite State Machine (FSM). The parameters of the queuing model are defined such that Service Level Agreements (SLA, e.g. as mean or percentiles of response times) are guaranteed except for overload conditions. The queuing model can be exactly analyzed under Markovian process assumptions from which all relevant quality of service (QoS) and energy efficiency (EE) metrics are derived. Numerical results are provided which demonstrate the applicability of the proposed model for the DC management, in particular to theoretically quantify the tradeoff between the conflicting aims of EE and QoS.

Deadline scheduling for aperiodic tasks in inter-Cloud environments: a new approach to resource management

In the big data era, the speed of analytical processing is influenced by the storage and retrieval capabilities to handle large amounts of data. While the distributed crunching applications themselves can yield useful information, the analysts face difficult challenges: they need to predict how much data to process and where, such that to get an optimum data crunching cost, while also respect deadlines and service level agreements within a limited budget. In today's data centers, data processing on demand and data transfers requests coming from distributed applications are usually expressed as aperiodic tasks. In this paper, we challenge the problem of tasks scheduling with deadline constraints of aperiodic tasks within inter-Cloud environments. In massively multithreaded computing systems that deal with data-intensive applications, Hadoop and BaTs tasks arrive periodically, which challenges traditional scheduling approaches previously proposed for supercomputing. Here, we consider the deadline as the main constraint, and propose a method to estimate the number of resources needed to schedule a set of aperiodic tasks, considering both execution and data transfers costs. Starting from classical scheduling techniques, and considering asynchronous tasks handling, we analyze the possibility of decoupling task arriving from task creation, scheduling and execution, sets of actions that can be put into a peer-to-peer relation over a network or over a client---server architecture in the Cloud. Based on a mathematical model, and using different simulation scenarios, we prove the following statements: (1) multiple source of independent aperiodic tasks can be considered similar to a single one; (2) with respect to the global deadline, the tasks migration between different regional centers is the appropriate solution when the number of estimated resources exceed a data center capacity; and (3) in a heterogeneous data center, we need a higher number of resources for the same request in order to respect the deadline constraints. We believe such results will benefit researchers and practitioners alike, who are interested in optimizing the resource management in data centers according to novel challenges coming from next-generation big data applications.

Hybrid shuffled frog leaping algorithm for energy-efficient dynamic consolidation of virtual machines in cloud data centers

Cloud computing aims to provide dynamic leasing of server capabilities as scalable virtualized services to end users. However, data centers hosting cloud applications consume vast amounts of electrical energy, thereby contributing to high operational costs and carbon footprints. Green cloud computing solutions that can not only minimize the operational costs but also reduce the environmental impact are necessary. This study focuses on the Infrastructure as a Service model, where custom virtual machines (VMs) are launched in appropriate servers available in a data center. A complete data center resource management scheme is presented in this paper. The scheme can not only ensure user quality of service (through service level agreements) but can also achieve maximum energy saving and green computing goals. Considering that the data center host is usually tens of thousands in size and that using an exact algorithm to solve the resource allocation problem is difficult, the modified shuffled frog leaping algorithm and improved extremal optimization are employed in this study to solve the dynamic allocation problem of VMs. Experimental results demonstrate that the proposed resource management scheme exhibits excellent performance in green cloud computing.

Novel Resource Allocation Algorithm for Energy-Efficient Cloud Computing in Heterogeneous Environment

Energy efficiency is one of the most important design considerations for a cloud data center. Recent approaches to the energy-efficient resource management for data centers usually model the problem as a bin packing problem with the goal of minimizing the number of physical machines (PMs) employed. However, minimizing the number of PMs may not necessarily minimize the energy consumption in a heterogeneous cloud environment. To address the problem, this paper models the resource allocation problem in a heterogeneous cloud data center as a constraint satisfaction problem (CSP). By solving this constraint satisfaction problem, an optimal resource allocation scheme, which includes a virtual machine provision algorithm and a virtual machine packing algorithm, is designed to minimize the energy consumption in a virtualized heterogeneous cloud data center. Performance studies show that this proposed new scheme outperforms the existing bin-packing based approaches in terms of energy consumption in heterogeneous cloud data centers.

Service level agreement based energy-efficient resource management in cloud data centers

As cloud computing has become a popular computing paradigm, many companies have begun to build increasing numbers of energy hungry data centers for hosting cloud computing applications. Thus, energy consumption is increasingly becoming a critical issue in cloud data centers. In this paper, we propose a dynamic resource management scheme which takes advantage of both dynamic voltage/frequency scaling and server consolidation to achieve energy efficiency and desired service level agreements in cloud data centers. The novelty of the proposed scheme is to integrate timing analysis, queuing theory, integer programming, and control theory techniques. Our experimental results indicate that, compared to a statically provisioned data center that runs at the maximum processor speed without utilizing the sleep state, the proposed resource management scheme can achieve up to 50.3% energy savings while satisfying response-time-based service level agreements with rapidly changing dynamic workloads.

Recovering Host Overloading in Cloud using Virtual Machines

provides an efficient solution to the objectives of the cloud computing paradigm by facilitating creation of Virtual Machines (VMs) over the underlying physical servers, leading to improved resource utilization and intellection. Virtualization proposes to create a virtual version of a device or a resource likely to virtualize a server, a storage space, operating system or even network where the mechanism divides the resource into one or more execution environments. To analyze the behavior of a Cloud data center by means of various aspects like availability, utilization, responsiveness and waiting time. The factors that a cloud provider must take into account are elasticity, scalability, live migration of VMs and performance isolation. Live migration of VMs, the process of dynamically transferring a virtual machine across different servers, hasproved to represent a new opportunity to enable responsive and dynamic proved torepresent a new opportunity toenable responsive and dynamic resource management in modern data centers. Migrations of VMs is performed on the basis of overloading that occurs in physical servers in cloud data centers, whichleads to relatively less performance degradation. The SLA violation is also less compared to other techniques and this means the cloud provider will incur less cost from VM migrations. Further, MHOD algorithm performs host selection and VM reallocation quicker than the existing algorithms. Maintenance of physical servers can be efficiently achieved through our algorithm as it leads to efficient consolidation of VMs.

Optimization and Mathematical Modeling in Computer Architecture

In the last few decades computer systems and the underlying hardware have steadily become larger and more complex. The need to increase their efficiency through architectural innovation has not abated, but quantitatively evaluating the effect of various choices has become more difficult. Performance and resource consumption are determined by complex interactions between many modules, each with many possible alternative implementations. We need powerful computer programs to explore large design spaces, but the traditional approach of developing simulators, building prototypes, or writing heuristic-based algorithms in traditional programming languages is often tedious and slow. Fortunately mathematical optimization has made great advances in theory, and many fast commercial and academic solvers are now available. In this book we motivate and describe the use of mathematical modeling, specifically optimization based on mixed integer linear programming (MILP) as a way to design and evaluate computer systems. The major advantage is that the architect or system software writer only needs to describe what the problem is, not how to find a good solution. This greatly speeds up their work and, as our case studies show, it can often lead to better solutions than the traditional approach. In this book we give an overview of modeling techniques used to describe computer systems to mathematical optimization tools. We give a brief introduction to various classes of mathematical optimization frameworks with special focus on mixed integer linear programming which provides a good balance between solver time and expressiveness. We present four detailed case studies -- instruction set customization, data center resource management, spatial architecture scheduling, and resource allocation in tiled architectures -- showing how MILP can be used and quantifying by how much it outperforms traditional design exploration techniques. This book should help a skilled systems designer to learn techniques for using MILP in their problems, and the skilled optimization expert to understand the types of computer systems problems that MILP can be applied to. Fully operational source code for the examples used in this book is provided through the NEOS System at www.neos-guide.org/content/computer-architecture Table of Contents: Acknowledgments / Introduction / An Overview of Optimization / Case Study: Instruction Set Customization / Case Study: Data Center Resource Management / Case Study: Spatial Architecture Scheduling / Case Study: Resource Allocation in Tiled Architectures / Conclusions / Bibliography / Authors' Biographies

DCworms – A tool for simulation of energy efficiency in distributed computing infrastructures

In the recent years, energy-efficiency of computing infrastructures has gained a great attention. For this reason, proper estimation and evaluation of energy that is required to execute data center workloads became an important research problem. In this paper we present a Data Center Workload and Resource Management Simulator (DCworms) which enables modeling and simulation of computing infrastructures to estimate their performance, energy consumption, and energy-efficiency metrics for diverse workloads and management policies. We discuss methods of power usage modeling available in the simulator. To this end, we compare results of simulations to measurements of real servers. To demonstrate DCworms capabilities we evaluate impact of several resource management policies on overall energy-efficiency of specific workloads executed on heterogeneous resources.