Large Data Centers Research Articles

A Review on Cloud Computing Architectures

Cloud computing has emerged to meet the requirements of large, Internet based and data intensive applications. Commonly clouds are implemented on large data centers consisting of thousands of servers. All application processing and resources are centralized in these data centers. As the number of users increases, this centralized approach produce bottlenecks and affect the quality of cloud services. This brings inconvenience to users. Fortunately centralized approach is not the only way to provide cloud services. Other possible architectures for cloud computing are federated approach and peer to peer approach. Federated cloud combines independent clouds and provides interoperability between them. P2P clouds are a low cost model for cloud computing. This paper reviews centralized, federated and peer to peer approaches to cloud computing.

Utilizing Massive Compute Resources in the Public Cloud for Complex Image Processing

This paper explores the deployment of complex image processing applications in a public cloud to make massive processing jobs practical for post-production budgets and timescales. It uses an illustrative example of an iterative motion-compensated super-resolution pixel reconstruction technique deployed in a cloud platform that can process what was an 8-week job on local hardware in a few hours using a public cloud. However, the technique is not without its pitfalls. Even large public cloud data centers have practical limits that cannot be ignored. Real-world examples range from 35-mm restoration for digital 3D re-release to the latest Hollywood 120 frames/sec, 4K, 3D high-dynamic-range content. Large-scale resources can be leveraged for otherwise impractical techniques, opening the door to a new breed of high-end media processing tools and making them accessible to post-production facilities.

Enhancing service capability with multiple finite capacity server queues in cloud data centers

Cloud computing took a step forward in the efficient use of hardware through virtualization technology. And as a result, cloud brings evident benefits for both users and providers. While users can acquire computational resources on-demand elastically, cloud vendors can also utilize maximally the investment costs for data centers infrastructure. In the Internet era, the number of appliances and services migrated to cloud environment increases exponentially. This leads to the expansion of data centers, which become bigger and bigger. Not just that these data centers must have the architecture with a high elasticity in order to serve the huge upsurge of tasks and balance the energy consumption. Although in recent times, many research works have dealt with finite capacity for single job queue in data centers, the multiple finite-capacity queues architecture receives less attention. In reality, the multiple queues architecture is widely used in large data centers. In this paper, we propose a novel three-state model for cloud servers. The model is deployed in both single and multiple finite capacity queues. We also bring forward several strategies to control multiple queues at the same time. This approach allows to reduce service waiting time for jobs and managing elastically the service capability for the whole system. We use CloudSim to simulate the cloud environment and to carry out the experiments in order to demonstrate the operability and effectiveness of the proposed method and strategies. The power consumption is also evaluated to provide insights into the system performance in respect of performance-energy trade-off.

Systematic Literature Review on Resource Allocation and Resource Scheduling in Cloud Computing

The objective the work is intend to highlight the key features and afford finest future directions in the research community of Resource Allocation, Resource Scheduling and Resource management from 2009 to 2016. Exemplifying how research on Resource Allocation, Resource Scheduling and Resource management has progressively increased in the past decade by inspecting articles, papers from scientific and standard publications. Survey materialized in three fold process. Firstly, investigate on the amalgamation of Resource Allocation, Resource Scheduling and then proceeded with Resource management. Secondly, we performed a structural analysis on different author’s prominent contributions in the form of tabulation by categories and graphical representation. Thirdly, huddle with conceptual similarity in the field and also impart a summary on all resource allocations. In cloud computing environments, there are two players: cloud providers and cloud users. On one hand, providers hold massive computing resources in their large datacenters and rent resources out to users on a per-usage basis. On the other hand, there are users who have applications with fluctuating loads and lease resources from providers to run their applications. Further, delivers conclusions by conferring future research directions in the field of cloud computing, such as reduce clouds early in the Internet, combining Resource Allocation, Resource Scheduling and Resource management rather than a Cloud model for providing high quality results, etc.

POWER EFFICIENT TASK SCHEDULING MECHANISM IN CLOUD ENVIRONMENT: A REVIEW

Cloud Computing is being used widely all over the world by many IT companies as it provides various benefits to the users like cost saving and ease of use. However, with the growing demands of users for computing services, cloud providers are encouraged to deploy large datacenters which consume very high amount of energy and also contribute to the increase in carbon dioxide emission in the environment. Therefore, we require to develop techniques which will help to get more environment friendly computing i.e. Green Cloud Computing. In this paper, we have reviewed the existing mechanisms and a new strategy to reduce the carbon dioxide emissions in federated Cloud ecosystems. More specifically, we propose a solution that allows providers to determine the best green destination to reduce the carbon dioxide emissions of the whole federated environment.

Triangular data privacy-preserving model for authenticating all key stakeholders in a cloud environment

Cloud computing is a relatively new paradigm that provides numerous advantages to service providers, developers, and customers with respect to flexibility, scalability, and availability at a lower cost. Motivated by these technical and economical advantages, many data owners outsource their data to centralized large data centers where the data are not only stored but also shared among multiple users. This method of data outsourcing brings many new security challenges for data integrity. There have been several mechanisms proposed lately that allow data owners to use a public verifier (e.g., a third-party auditor (TPA)) for efficiently auditing cloud data integrity. The use of a TPA for this purpose is inevitable, since it provides several advantages to both cloud service users (CSUs) and cloud service providers (CSPs) in terms of efficiency, fairness, trust, etc. – which is essential to achieve economies of scale for cloud computing. Although the existing public auditing schemes are capable of simultaneously performing multiple auditing tasks (including the integrity of cloud data) in an efficient manner, these methods can reveal confidential information to public verifiers which makes a TPA a potential threat to the data security of CSUs and the reputation of CSPs. Therefore, cloud computing requires a holistic approach to security that can deal with all the potential threats that exist in the cloud environment. Taking these points into account, this paper presents a novel triangular data privacy-preserving (TDPP) model that supports public auditing with the capability of auditing all the key stakeholders (i.e., CSU, TPA, and CSP) for achieving optimal security in a cloud environment. Specifically, our proposed TDPP model supports three types of auditing. First, the TPA can audit the CSP for ensuring (a) the correctness of the CSU's data stored at the service provider side and (b) that the CSP is in full compliance of its own service level agreement (SLA). Second, the TPA can audit the CSU to (a) determine any violation of terms and conditions defined in the SLA for the cloud services provided by the CSP and (b) monitor the feedback provided by the CSU for the utilized services. Third, either the CSU or CSP can audit the TPA to minimize the possibility of any potential insider threats or attacks. For instance, the CSU can audit the TPA to (a) ensure that private information being shared with the TPA is not disclosed to any third party or misused by an insider and (b) verify that the TPA performs assigned auditing tasks according to the given specification within the agreed time frame. Our experimental results demonstrate the effectiveness and efficiency of our proposed scheme when auditing all key stakeholders.

Methodology for energy aware adaptive management of virtualized data centers

This paper proposes a methodology for energy aware management of virtualized data centers (DC) based on dynamically adapting and scaling the computing capacity to the characteristics of the workload. To assess the energy efficiency of DC operation, we have defined a novel ontological model for representing its energy and performance characteristics and a new metric for aggregating Green and Key Performance Indicators and calculating at run-time the DC Greenness Level. Workload balancing and consolidation is achieved by means of an automated reinforcement learning-based decision process targeting to increase the workload density and to scale down the unused computing resources. Evaluation results show that up to 15.6 % energy savings are obtained on our test bed DC. Tests conducted in a simulated environment show that the time and space overhead of our methodology are within reasonable limits and that by organizing the servers in hierarchical clusters, the methodology can manage highly dynamic workload in large DCs with thousands of servers. The methodology is already implemented in the Green Cloud Scheduler, an official component of the OpenNebula Middleware which is available in the OpenNebula Ecosystem web site to be downloaded and used.

How much power does your server consume? Estimating wall socket power using RAPL measurements

Full system electricity intake from the wall socket is important for understanding and budgeting the power consumption of large scale data centers. Measuring full system power, however, requires extra instrumentation with external physical devices, which is not only cumbersome, but also expensive and time consuming. To tackle this problem, in this paper, we propose to model wall socket power from processor package power obtained from the running average power limit (RAPL) interface, which is available on the latest Intel processors. Our experimental results demonstrate a strong correlation between RAPL package power and wall socket power consumption. Based on the observations, we propose an empirical power model to predict the full system power. We verify the model using multiple synthetic benchmarks (Stress-ng, STREAM), high energy physics benchmark (ParFullCMS), and non-trivial application benchmarks (Parsec). Experimental results show that the prediction model achieves good accuracy, which is maximum 5.6 % error rate.

A self‐organized load balancing mechanism for cloud computing

SummaryThe growth in computer and networking technologies over the past decades established cloud computing as a new paradigm in information technology. The cloud computing promises to deliver cost‐effective services by running workloads in a large scale data center consisting of thousands of virtualized servers. The main challenge with a cloud platform is its unpredictable performance. A possible solution to this challenge could be load balancing mechanism that aims to distribute the workload across the servers of the data center effectively. In this paper, we present a distributed and scalable load balancing mechanism for cloud computing using game theory. The mechanism is self‐organized and depends only on the local information for the load balancing. We proved that our mechanism converges and its inefficiency is bounded. Simulation results show that the generated placement of workload on servers provides an efficient, scalable, and reliable load balancing scheme for the cloud data center. Copyright © 2016 John Wiley & Sons, Ltd.

Dynamo

Data center power is a scarce resource that often goes underutilized due to conservative planning. This is because the penalty for overloading the data center power delivery hierarchy and tripping a circuit breaker is very high, potentially causing long service outages. Recently, dynamic server power capping, which limits the amount of power consumed by a server, has been proposed and studied as a way to reduce this penalty, enabling more aggressive utilization of provisioned data center power. However, no real at-scale solution for data center-wide power monitoring and control has been presented in the literature. In this paper, we describe Dynamo -- a data center-wide power management system that monitors the entire power hierarchy and makes coordinated control decisions to safely and efficiently use provisioned data center power. Dynamo has been developed and deployed across all of Facebook's data centers for the past three years. Our key insight is that in real-world data centers, different power and performance constraints at different levels in the power hierarchy necessitate coordinated data center-wide power management. We make three main contributions. First, to understand the design space of Dynamo, we provide a characterization of power variation in data centers running a diverse set of modern workloads. This characterization uses fine-grained power samples from tens of thousands of servers and spanning a period of over six months. Second, we present the detailed design of Dynamo. Our design addresses several key issues not addressed by previous simulation-based studies. Third, the proposed techniques and design have been deployed and evaluated in large scale data centers serving billions of users. We present production results showing that Dynamo has prevented 18 potential power outages in the past 6 months due to unexpected power surges; that Dynamo enables optimizations leading to a 13% performance boost for a production Hadoop cluster and a nearly 40% performance increase for a search cluster; and that Dynamo has already enabled an 8% increase in the power capacity utilization of one of our data centers with more aggressive power subscription measures underway.

Costly Circuits, Submodular Schedules and Approximate Carathéodory Theorems

Hybrid switching -- in which a high bandwidth circuit switch (optical or wireless) is used in conjunction with a low bandwidth packet switch -- is a promising alternative to interconnect servers in today's large scale data centers. Circuit switches offer a very high link rate, but incur a non-trivial reconfiguration delay which makes their scheduling challenging. In this paper, we demonstrate a lightweight, simple and nearly-optimal scheduling algorithm that trades-off reconfiguration costs with the benefits of reconfiguration that match the traffic demands. Seen alternatively, the algorithm provides a fast and approximate solution towards a constructive version of Caratheodory's Theorem for the Birkhoff polytope. The algorithm also has strong connections to submodular optimization, achieves a performance at least half that of the optimal schedule and strictly outperforms state of the art in a variety of traffic demand settings. These ideas naturally generalize: we see that indirect routing leads to exponential connectivity; this is another phenomenon of the power of multi-hop routing, distinct from the well-known load balancing effects.

Power‐aware server consolidation for federated clouds

SummaryCloud computing has evolved to provide computing resources on‐demand through a virtualized infrastructure, letting applications, computing power, data storage, and network resources to be provisioned and managed over private networks or over the Internet. Cloud services normally run on large data centers and demand a huge amount of electricity. Consequently, the electricity cost represents one of the major concerns of data centers, because it is sometimes nonlinear with the capacity of the data centers, and it is also associated with a high amount of carbon emission (CO2). However, energy‐saving schemes that result in too much degradation of the system performance or in violations of service‐level agreement (SLA) parameters would eventually cause the users to move to another cloud provider. Thus, there is a need to reach a balance between energy savings and the costs incurred by these savings in the execution of the applications. Therefore, in this paper, we propose and evaluate a power and SLA‐aware application consolidation solution for cloud federations. It comprises a multi‐agent system for server consolidation, taking into account SLA, power consumption, and carbon footprint. Different for similar solutions available in the literature, in our solution, when a cloud is overloaded, its data center needs to negotiate with other data centers before migrating the workload to another cloud. Simulation results show that our approach can reduce up to 46% of the power consumption while trying to meet performance requirements. Furthermore, we show that federated clouds can provide an adequate solution to deal with power consumption in the clouds. Copyright © 2016 John Wiley & Sons, Ltd.

Effects of structural parameters on fluid flow and heat transfer characteristics in microchannel with offset zigzag grooves in sidewall

With the rapid development of large data center, the required power consumption increases sharply. The thermal management of that accounts for about 33% of energy consumption. Single phase liquid convection heat transfer of microchannel heat sink has been thought as one of the most promising methods. For energy efficient cooling, optimized microchannel structures is one important approach. Hence, the novel offset zigzag microchannnel is proposed and optimized for inlet Reynolds number ranging from 200–800 in this paper. The heat sink is composed of 30 parallel channels, and basic structure parameters of rectangular channel are width of 0.1mm, depth of 0.3mm, length of 5mm and pitch of 0.2mm. The numerical model considers entrance effect, viscous heating and the temperature-dependent thermal-physical properties. In order to study the influence of geometry on flow and heat transfer characteristics, two non-dimensional variables are defined, such as the length ratio of zigzag length to channel length (α) and the length ratio of zigzag expansion to zigzag length (γ). Results indicate the zigzag channel can effectively prevent the rise of heat surface temperature along the flow direction and increase the minimum temperature, which improves the temperature distribution uniformity. Under Rein=650 and α=0.04, the minimum temperature increases from 309.9K to 310.1K and maximum temperature decreases from 324.02K to 318.91K. For 0.04⩽α<0.1 and α<0.04 of Rein<500, the thermal characteristic is enhanced and the flow resistance also is reduced. Compared with rectangle channel, the maximum temperature and pressure drop of zigzag channel with α=0.04 of Rein=650 are also reduced from 324.02K to 318.76K and 29.81kPa to 26.75kPa, respectively. It can be interpreted that with the decrease of α, the porosity and convection heat transfer areas are increased and turbulence intensity of fluid flow is enhanced. The smaller γ is more favorable for the enhancement of heat transfer except for γ=0 due to the fluid stagnation at the bottom of zigzag cavity. However, compared with rectangular microchannel, heat transfer is enhanced for all cases and flow resistance is decreased besides γ=0.1 of Rein>500. For γ=0 of Rein=650, the maximum temperature and pressure drop are reduced from 324.02K to 318.01K and 298.81kPa to 27.91kPa, respectively. The overall thermal performance is measured with total thermal resistance vs pumping power. The best performance of offset zigzag microchannel is found at α=0.04 and γ=0.1. For Rein=800, the maximum temperature of heat surface is decreased 5.65K, the minimum increased 0.36K, and pumping power is decreased 1.4%. And thermal resistance is reduced by 17.4% at the Pp=0.167W.

Evolutionary Computing Based on QoS Oriented Energy Efficient VM Consolidation Scheme for Large Scale Cloud Data Centers

Abstract The high pace and increase in cloud computing technology and associated applications, especially large scale data centres, have demanded energy efficient and Quality of Service (QoS) oriented computing platform. To meet these requirements, virtualization and Virtual Machine (VM) consolidation has emerged as an effective solution. The optimization in VM consolidation by means of efficient dynamic resource-utilization prediction, VM selection and placement can achieve optimal solution for energy efficient and QoS oriented cloud computing system. In this paper, an evolutionary computing algorithm called Adaptive Genetic Algorithm (A-GA) based VM consolidation approach has been developed. A-GA based placement policy and its implementation with different VM selection policies like Minimum Migration Time (MMT), Maximum Correlation (MC) and Random Selection (RS), along with different CPU utilization estimation approaches like Inter Quartile Range (IQR), Local Regression (LR), Local Robust Regression (LRR), static THReshold (THR) and Median Absolute Deviation (MAD) has revealed that A-GA based consolidation with MMT selection policy and combined IQR and LRR can enable optimal VM consolidation for large scale infrastructures. In addition, the proposed A-GA policy has exhibited better performance as compared to other meta-heuristics such as Ant Colony Optimization (ACO) and Best Fit Decreasing. The proposed consolidation system can be used for large scale cloud infrastructures where energy conservation, minimal Service Level Agreement (SLA) violation and QoS assurance is inevitable.

Joint Static and Dynamic Traffic Scheduling in Data Center Networks

The advent and continued growth of large data centers has led to much interest in switch architectures that can economically meet the high capacities needed for interconnecting the thousands of servers in these data centers. Various multilayer architectures employing thousands of switches have been proposed in the literature. We make use of the observation that the traffic in a data center is a mixture of relatively static and rapidly fluctuating components, and develop a combined scheduler for both these components using a generalization of the load-balanced scheduler. The presence of the known static component introduces asymmetries in the ingress-egress capacities, which preclude the use of a load-balanced scheduler as is. We generalize the load-balanced scheduler and also incorporate an opportunistic scheduler that sends traffic on a direct path when feasible to enhance the overall switch throughput. Our evaluations show that this scheduler works very well despite avoiding the use of a central scheduler for making packet-by-packet scheduling decisions.

A Classification-Based Virtual Machine Placement Algorithm in Mobile Cloud Computing

In recent years, cloud computing services based on smart phones and other mobile terminals have been a rapid development. Cloud computing has the advantages of mass storage capacity and high-speed computing power, and it can meet the needs of different types of users, and under the background, mobile cloud computing (MCC) is now booming. In this paper, we have put forward a new classification-based virtual machine placement (CBVMP) algorithm for MCC, and it aims at improving the efficiency of virtual machine (VM) allocation and the disequilibrium utilization of underlying physical resources in large cloud data center. By simulation experiments based on CloudSim cloud platform, the experimental results show that the new algorithm can improve the efficiency of the VM placement and the utilization rate of underlying physical resources.

A Framework for Block Placement, Migration, and Fast Searching in Tiled-DNUCA Architecture

Multicore processors have proliferated several domains ranging from small-scale embedded systems to large data centers, making tiled CMPs (TCMPs) the essential next-generation scalable architecture. NUCA architectures help in managing the capacity and access time for such larger cache designs. It divides the last-level cache (LLC) into multiple banks connected through an on-chip network. Static NUCA (SNUCA) has a fixed address mapping policy, whereas dynamic NUCA (DNUCA) allows blocks to relocate nearer to the processing cores at runtime. To allow this, DNUCA divides the banks into multiple banksets and a block can be placed in any bank within a particular bankset. The entire bankset may need to be searched to access a block. Optimal bankset searching mechanisms are essential for getting the benefits from DNUCA. This article proposes a DNUCA-based TCMP architecture called TLD-NUCA. It reduces the LLC access time of TCMP and also allows a heavily loaded bank to distribute its load among the underused banks. Instead of other DNUCA designs, TLD-NUCA considers larger banksets. Such relaxations result in more uniform load distribution than existing DNUCA-based TCMP (T-DNUCA). Considering larger banksets improves the utilization factor, but T-DNUCA cannot implement it because of its expensive searching mechanism. TLD-NUCA uses a centralized directory, called TLD, to search a block from all the banks. Also, the proposed block placement policy reduces the instances when the central TLD needs to be contacted. It does not require the expensive simultaneous search as needed by T-DNUCA. Better cache utilization and a reduction in LLC access time improve the miss rate as well as the average memory access time (AMAT). Improving the miss rate and AMAT results in improvements in cycles per instructions (CPI). Experimental analysis found that TLD-NUCA improves performance by 6.5% as compared to T-DNUCA. The improvement is 13% as compared to the SNUCA-based TCMP design.

Empirical comparison of power-efficient virtual machine assignment algorithms

The advent of cloud computing has changed the way many companies do computation, allowing them to outsource it to the cloud. This has given origin to a new kind of business, the cloud providers, which run large datacenters. In order to be competitive, cloud providers must keep their operational costs low. One way to reduce these costs is reducing the energy consumed with smart task assignment algorithms, which decide where tasks are to be placed upon their arrival. Unfortunately, almost no task assignment algorithm used is power aware.In this paper we compare the performance of multiple task assignment algorithms for saving energy. We assume that tasks are in fact virtual machines that have to be assigned to physical machines, and we assume that the physical machines have a power consumption that increases superlinearly with the load. First, we propose two tunable power-aware task assignment algorithms (that subsume the algorithms studied in Arjona et al., Power-efficient assignment of virtual machines to physical machines, Future Generation Computer Systems, 2015). These algorithms are then compared with multiple state-of-the-art algorithms in different meaningful scenarios. Both algorithms prove themselves as interesting assignment algorithms since, properly configured, they outperform the other algorithms in most of the cases.

Optimization of Data Center Battery Storage Investments for Microgrid Cost Savings, Emissions Reduction, and Reliability Enhancement

This paper presents a methodology for optimizing investment in data center battery storage capacity. Utility grid managers spend significant resources toward predicting and matching available power generation capacity to demand in real time. It is therefore essential for the success of the power industry that economic dispatch, energy efficiency, and grid security be maintained as power requirements change. This is especially challenging for microgrids during periods of peak demand due to limited available capacity. Data centers possess a unique requirement for short-term battery power supply where cost savings, emissions reduction, and reliability enhancement can be achieved through investment in additional battery capacity. To maximize these benefits, an optimization methodology is presented through a case study for an existing data center and microgrid. Here, we discuss a case study demonstrating the effectiveness of the proposed approach. For the selected mid-size data center, our results indicate monetize monthly savings of up to 10 000$ and 0.5% reduction in loss of load probability while simultaneously reducing carbon footprints. The results of this work are directed toward large data centers at university and corporate campuses, microgrids, and military installations.

A study on the evaluation of ventilation system suitable for outside air cooling applied in large data center for energy conservation

In developed countries, expansion of communication technology has resulted in continual increase in the construction of data centers with high-density cooling loads. Throughout a year, IT equipment installed in a data center generates large and constant cooling load. As a result, data centers may be consuming an ever-growing amount of energy. The cooling system utilizing the energy of outside air is applied universally to reduce data center energy consumption. The application of the cooling system to the outdoor air cooling system of a data center considers that temperature efficiency and ventilation performance vary depending on the type of ventilation system. The displacement and mixed ventilation method can be applied generally to a data center. The efficiency of a ventilation system depends on inside temperature or contaminant concentrations in room and outlets. This study thus aims to evaluate the ventilation performance that varies according to type of ventilation system installed in the data center. Ventilation efficiency is assessed by applying the concept of total air age and considers the fresh air ratio and age of return air. Further, temperature efficiency gained by utilizing temperature difference is used to assess causes for changes in ventilation performance.