Search

In edge computing, application components can be placed over a range of computational devices from cloud data centers to nodes at the network edge. Application placement can have significant impact on important metrics like latency and resource utilization. Thus, application placement is an important optimization problem. In edge computing, the characteristics of both the infrastructure and the application may change over time, which may require the dynamic re-optimization of the application placement. Most algorithms suggested so far for the dynamic re-optimization of edge application placement are centralized, i.e., they rely on one entity collecting information from the whole infrastructure and making decisions centrally. However, centralized approaches suffer from limited scalability and are vulnerable to failures. In this paper, we present a decentralized approach for the dynamic re-optimization of edge application placement. We adopt an algorithm of Malek et al. for distributed systems and modify it to make it applicable to edge computing. In this approach, each node makes decisions autonomously, using auctions for coordination. Our empirical results demonstrate that the proposed algorithm is very effective in optimizing edge application placement. In an edge system with 637 edge nodes and 563 end devices, our algorithm achieves 54% higher reduction of application latency than a previous decentralized algorithm. • Optimized placement of application components on cloud and edge devices. • Decentralized optimization algorithm using local knowledge and local modifications. • Auctions and distributed locking for synchronization. • Minimization of application latency. [ABSTRACT FROM AUTHOR]

Scheduling independent tasks on a parallel platform is a widely-studied problem, in particular when the goal is to minimize the total execution time, or makespan (P | | C max problem in Graham's notations). Also, many applications do not consist of sequential tasks, but rather parallel tasks, either rigid, with a fixed degree of parallelism, or moldable, with a variable degree of parallelism (i.e., for which we can decide at the execution on how many processors they are executed). Furthermore, since the energy consumption of data centers is a growing concern, both from an environmental and economical point of view, minimizing the energy consumption of a schedule is a main challenge to be addressed. One can then decide, for each task, on how many processors it is executed, and at which speed the processors are operated, with the goal to minimize the total energy consumption. We further focus on co-schedules, where tasks are partitioned into shelves, and we prove that the problem of minimizing the energy consumption remains NP-complete when static energy is consumed during the whole duration of the application. We are however able to provide an optimal algorithm for the schedule within one shelf, i.e., for a set of tasks that start at the same time. Several approximation results are derived, both with discrete and continuous speed models, and extensive simulations are performed to show the performance of the proposed algorithms. • A model for a theoretical analysis of the energy consumption of a schedule for moldable and rigid tasks. • Most presented scheduling problems are NP-complete. • There exist approximation algorithms for the presented problems (from 1.59 to 3 depending on the problem). • An experimental study of the trade-off between the time complexity of the scheduler and the energy consumed by the schedule. [ABSTRACT FROM AUTHOR]

Energy is a considerable portion of the cost of a cloud data center. In addition to energy, carbon emission tax imposes another cost on data centers, which have different policies in different cities around the world. Most of the cloud data centers are geographically distributed, which can have a huge advantage in reducing such costs. In addition, in recent years, cloud federation, in which multiple cloud providers share their IT infrastructures voluntarily, could play a crucial role in minimizing the energy consumption of cloud data centers. On the other hand, improper VM placement results frequent VM migration, constant turning off/on physical machines, service quality degradation and energy consumption increase. To tackle these weaknesses, we propose a multi-objective algorithm, Called OUR-ACS, to minimize data centers' energy consumption, carbon emission, and the total expenses (energy cost + carbon tax) in a federated environment considering both initial VM placement and VM consolidation approaches. We evaluate the efficiency of our proposed algorithm by using the ClousSim Plus simulator toolkit using real-world datasets. Compared to the competing algorithms, our simulation results indicate that our proposed algorithm could reduce the total energy consumption, carbon emission, and the total costs of multiple cloud data centers by an average of 37.6%, 41%, and 25%, respectively, in a federated environment. • We propose a novel multi-objective cloud federation algorithm considering to minimize three objectives. • We have identified a new problem in VM migration called Migration lock-in status. • We employ new heuristics to make a tradeoff between servers' power consumption and their resource utilization simultaneously. • We conduct a comprehensive simulation under an IaaS cloud infrastructure using a real-world datasets. [ABSTRACT FROM AUTHOR]

The diverse needs of customers drive cloud providers to incorporate more GPU-enabled services. It is known that users barely utilize GPUs. Hence, GPU virtualization techniques are employed to enable GPU sharing and increase resource utilization. Among GPU virtualization techniques, API remoting leverages disaggregation for GPU provisioning, allowing GPUs to be accessed remotely by non-resident VMs. It enables a VM to access any available GPU in the data center which, in turn, provides more flexibility to reach a better placement. It also enables users to choose VM and vGPU instances separately based on computational needs as GPU-enabled VM instances with fixed configurations may not proportionally utilize allocated resources. However, an inefficient VM or vGPU placement may result in poor performance and increase in energy consumption and rejection ratio. There is also a network communication overhead for remotely allocated vGPUs. The main challenge is how to place VMs and vGPUs such that rejection ratio is decreased and the negative performance impacts caused by GPU sharing and remote access are kept at a minimum. In this work, we formally define the joint problem of VM and vGPU placement based on API remoting as a multi-objective ILP model and introduce gVMP as a heuristic to solve it. The proposed method is compared against state-of-the-art heuristics and commercial solutions that are based on API remoting and GRID technologies for the placement of GPU-enabled VMs. We study the effectiveness of the proposed gVMP algorithm for low and high arrival rates in slow and fast networks. Our results show that under different scenarios and in comparison to state-of-the-art policies, our heuristic improves average request duration, efficiency, and SLA by up to 40%, 25% and 29%, respectively. • The joint problem of VM and vGPU placement based on API remoting is formulated as a multi-objective ILP model. • The proposed heuristic reduces rejection ratio and negative performance impacts of GPU sharing and remote vGPU accesses. • The results are compared against state-of-the-art and commercial solutions for API remoting and GRID technologies. • The effectiveness is studied for low and high arrival rates in slow and fast networks under different scenarios. • The heuristic improves average request duration, efficiency, and SLA by up to 40%, 25% and 29%, respectively. [ABSTRACT FROM AUTHOR]

The workload of Cloud data centers is constantly fluctuating causing imbalances across physical hosts that may lead to violations of service-level agreements. To mitigate workload imbalances, this work proposes a concurrent agent-based problem-solving technique supported by cooperative game theory capable of balancing workloads by means of live migration of virtual machines (VMs). Nearby agents managing physical hosts are partitioned into coalitions in which agents play coalitional games to progressively balance separate sections of a data center while considering the coalition's benefit of migrating a VM as well as its associated network overhead. Simulation results show that, in general, the proposed coalition-based load balancing mechanism outperformed a load balancing mechanism based on a hill-climbing algorithm used by a top data center vendor when considering altogether (i) the standard deviation of resource usage, (ii) the number of migrations, and (iii) the number of switch hops per migration. • Agents in coalitions progressively balance data center sections. • Supported by a game-theoretical framework, agents cooperate to balance workloads. • Coalitions composed of nearby hosts are used as a basis to balance workloads. • Coalition-aware heuristics consider both the benefit of migrating a VM and its cost. [ABSTRACT FROM AUTHOR]

Reducing energy consumption and optimizing resource usage in large cloud data centers is still an essential target for the current researchers and cloud providers. The state-of-the-art highlights the effectiveness of VM consolidation and live migrations in achieving reasonable solutions. However, most proposals consider only the real-time workload variations to decide whether a host is overloaded or underloaded, or to trigger migration actions. Such approaches may apply frequent and needless VM migrations leading to energy waste, performance degradation, and service-level agreement (SLA) violations. In this paper, we propose a consolidation approach based on the resource utilization prediction to determine the overloaded and underloaded hosts. The prediction method combines a Kalman filter and support vector regression (SVR) to forecast the host's future CPU utilization. Simulations are conducted on Cloudsim using real PlanetLab workloads to verify the performance of our proposal against existing benchmark algorithms. Experimental results demonstrate that our consolidation technique significantly reduces the SLA violation rate, number of VM migrations, and energy consumed in the datacenter. • A multi-step workload prediction model based on Kalman Filter and SVR is proposed. • A dynamic VM consolidation approach based on resource prediction is presented. • A trade-off between SLA violation and energy consumption is pursued. • Analysis study is carried out on the energy consumed by executing our simulation. • Several experiments are conducted on Cloudsim to evaluate our approach. [ABSTRACT FROM AUTHOR]

• An GAN-based method for time-dependent cloud workload generation is proposed. • The method does not rely on any prior knowledge to capture the data distribution. • Spectral normalization is used to stabilize the adversarial training. • Achieving better results than state-of-art models. • Enabled conditional workload generation with simple changes on models. To design repeatable and comparable resource management policies for data centers, researchers mainly conduct experiments in the simulation environment, which requires large-scale workload traces to simulate real scenes. However, issues related to data collection, security and privacy hinder the public availability of cloud workload datasets. Though workload generation is a promising solution, due to the unpredictable time dependency, cloud workloads are difficult to model. In light of this, we propose a novel end-to-end model for time-dependent cloud workload generation using Generative Adversarial Networks, which adopts improved Temporal Convolution Networks and Spectral Normalization to capture the time dependency and stabilize the adversarial training. Experimental results on real cloud datasets demonstrate that our model can efficiently generate realistic workloads that fulfill the diversity, fidelity and usefulness. Further, we also propose a conditional GAN which is trained with labeled data and can generate specific kind of workloads according to the input. [ABSTRACT FROM AUTHOR]

The erasure-coded cross-data center storage system can achieve high disaster tolerance and low redundancy. But as it has large cross-data center update traffic, its data update time is long. In erasure-coded storage systems, each data object is sequentially divided into several stripes, and each stripe consists of several data packets. When erasure-coded stripes do not undergo insertion or deletion, existing work can effectively reduce cross-data center update traffic by performing delta updates—delta update methods can update the old stripe without transferring matched packets that are new and old stripes' duplicate packets with the same offset-within-stripe. However, because existing delta update methods' stripe size is fixed, when a stripe undergoes insertion or deletion, its subsequent stripes' duplicate packets' offset-within-stripe will change. In this scenery, the matched packet number is small, resulting in large cross-data center update traffic. This paper proposes an elastic stripe-based delta update method for erasure-coded cross-data center storage systems (ESDU). Under insertion or deletion, ESDS tries to avoid duplicate packets' offset-within-stripe changing (i.e., maximizing matched packets) by adjusting the stripes' size flexibly according to the duplicate packet locating result. So, it can reduce cross-data center traffic. Moreover, ESDU can optimize stripes' update topology based on the location information of storage nodes to reduce cross-data center update traffic further. In addition, we implement an erasure-coded cross-data center storage system adopting ESDU, called ECESD. Experiments with the workloads derived from EduCoder's real-world trace show that compared with the existing erasure-coded cross-data center storage system adopting the fixed stripe-based delta update method, ECESD reduces average update time by 89.6%. Moreover, compared with a replication-based storage system with a delta update method (HadoopRsync), ECESD achieves an 8.3% shorter update time and much smaller redundancy. • We propose an elastic stripe-based delta update method for erasure-coded cross-data center storage systems, ESDU. • ESDU can shorten data update time by adjusting stripe size and optimizing update topology reasonably. • We implement an erasure-coded cross-data center storage system adopting ESDU called ECESD. • ECESD reduces average update time by 89.6%, compare with existing erasure-coded cross-data center storage systems. • ECESD's average update time is even shorter than existing replication-based cross-data center storage systems. [ABSTRACT FROM AUTHOR]

The vigorous development and the increasing popularity of cloud computing highlight the necessity of reducing data center energy consumption and the environmental impact of carbon dioxide emissions. For geographically distributed data centers, cloud servers are connected to the conventional power grid and in addition they are supported by an attached renewable energy source. Since the carbon footprint rate of energy consumption has dynamic differences in space, reducing energy consumption does not mean decrease carbon emission, which indicates that energy consumption and carbon footprint need to be synergistically optimized. In this paper, an energy and carbon-aware algorithm for virtual machine placement is proposed. The goal is to obtain a virtual machine allocation scheme that aims to achieve the trade-off between energy consumption and carbon emissions by improving renewable energy utilization. The experimental results show that the proposed approach is more energy-efficient and greener, which can also maximize the renewable energy utilization with 73.11% while ensuring the SLA violation with 0.2% in comparison to the baseline algorithms. • Propose a novel energy-aware VM placement approach in cloud data centers. • Introduce renewable energy for managing energy consumption and carbon footprint. • Design renewable energy generation prediction method based on EEMD and TCN. • Demonstrate the superiority of the proposed algorithm with five baseline algorithms. [ABSTRACT FROM AUTHOR]

• A deployment strategy for edge nodes based on the analysis of mobility traces. • A performance comparison of the strategy against traditional approaches. • The use of ParticpAct and Google Popular Time. • A data exploration analysis of the data sets describing some mobility features. The main objective of Multi-access Edge Computing (MEC) is to bring computational capabilities at the edge of the network to better support low-latency applications. Such capabilities are typically offered by Edge Data Centers (EDC). The MEC paradigm is not tied to a single radio technology, rather it embraces both cellular and other radio access technologies such as WiFi. Distributed intelligence at the edge for AI purposes requires careful spatial planning of computing and storage resources. The problem of EDC deployment in urban environments is challenging and, to the best of our knowledge, it has been explored only for cellular connectivity so far. In this paper, we study the possibility of deploying EDC without analyzing the expected data traffic load of the cellular network, a kind of information rarely shared by network operators. To this purpose, we propose in this work CLUB, CLUstering-Based strategy tailored on the analysis of urban mobility. We analyze two experimental mobility data sets, and we analyze some mobility features in order to characterize their properties. Finally, we compare the performance of CLUB against state-of-the-art techniques in terms of the outage probability, namely the probability an EDC is not able to serve a request. Our results show that the CLUB strategy is always comparable with respect to our benchmarks, but without using any information related to network traffic. [ABSTRACT FROM AUTHOR]

• We propose a new approach to achieve balance and optimization in resource exploitation and reduce energy consumption in the Cloud. • We propose two successive algorithms of FCA and Fuzzy-FCA to adapt our data according to their type (exact or fuzzy). • We use a mixed environment by integrating the MEC architecture with the cloud to minimize end-to-end latency. • We offer a feasible solution to properly place network services at VNFs. • We apply Fuzzy FCA to place and consolidate the VNFs based on a classification very effective. Virtual Network Function (VNF) is one of the pillars of a Cloud network that separates network functions and their dedicated hardware devices, such as routers, firewalls, and load balancers, to host their services on virtual machines. The VNF is responsible for network services that run on virtual machines and can connect each of them alone or organize themselves into a single enclosure to use all the resources available in that enclosure. This flexibility allows physical and virtual resources to be used in a way that ensures control over power consumption, balance in resource use, and minimizing costs and latency. In order to consolidate VNF groups into a minimum number of Virtual Machine (VM) with estimation of the association relation to a measure of confidence under the context of possibility theory, we propose a new Fuzzy-FCA approach for VNF placement based on Formal Concept Analysis (FCA) and fuzzy logic in mixed environment based on cloud data centers and Multiple access Edge Computing (MEC) architecture. Thus, the inclusion of this architecture in the cloud environment ensures the distribution of compute resources to the end user in order to reduce end-to-end latency. To confirm the effectiveness of our solution, we compared it to one of the best algorithms studied in the literature, namely the MultiSwarm algorithm. The results of the series of experiments carried out show the feasibility and efficiency of our algorithm. Indeed, the harvested results confirm the capability of maximizing and balancing the use of resources, of minimizing the latency and the cost of energy consumption. The performance of our solution in terms of average latency represents 16%, a slight increase compared to MultiSwarm, and an average gain, in runtime, of 49%, compared to the same algorithm. [ABSTRACT FROM AUTHOR]

• FPGA performance characterization using parametrizable OpenCL benchmarks. • High resource utilization and performance on Intel and Xilinx FPGAs. • Ability to also measure the performance of devices and tools. Emerging high-level tools lead to a reduced development time for applications on FPGA accelerators while still producing high-quality results. This is one reason for the increased adoption of FPGAs in data center applications which emphasizes the need for a benchmark suite to enable the comparison of FPGA architecture, programming tools, runtimes, and libraries. Because of the lack of such a benchmark suite, we have developed an OpenCL-based open-source implementation of the HPCC benchmark suite for Xilinx and Intel FPGAs. In an in-depth evaluation, we show that the benchmarks allow to quantify the impact of HBM2 memory in comparison to FPGAs with DDR and to analyze differences in the arithmetic units on current FPGA architectures. Power measurements indicate, that not all benchmark implementations can utilize the full potential of the FPGAs in terms of power efficiency. We are continuing to optimize and port the benchmark for new generations of FPGAs and design tools and we encourage active participation to create a valuable tool for the community. [ABSTRACT FROM AUTHOR]

The k -ary n -cube Q n k serves as an indispensable interconnection network in the design of data center networks, network-on-chips, and parallel computing systems since it possesses numerous attractive properties. In these parallel architectures, the paired (or unpaired) many-to-many m -disjoint path cover (m -DPC) plays a significant role in message transmission. Nevertheless, the construction of m -DPC is severely obstructed by large-scale edge faults due to the rapid growth of the system scale. In this paper, we investigate the existence of paired 2-DPC in Q n k under the partitioned edge fault (PEF) model, which is a novel fault model for enhancing the networks' fault-tolerance related to path embedding problem. We exploit this model to evaluate the edge fault-tolerance of Q n k when a paired 2-DPC is embedded into Q n k. Compared to the other known works, our results can help Q n k to achieve large-scale edge fault-tolerance. • This research studies the existence of paired 2-disjoint path cover in k -ary n -cubes in the presence of large-scale faulty edges. • Our findings generalize the result in [50] from the theoretical perspective. • Our results provide the exponential edge fault-tolerance compared to previous similar works. [ABSTRACT FROM AUTHOR]

The increasing problem of bacterial resistance, particularly with quinolone-resistant Escherichia coli (QnR eco) poses a serious global health issue. We collected data on QnR eco resistance rates and detection frequencies from 2014 to 2021 via the China Antimicrobial Resistance Surveillance System, complemented by meteorological and socioeconomic data from the China Statistical Yearbook and the China Meteorological Data Service Centre (CMDC). Comprehensive nonparametric testing and multivariate regression models were used in the analysis. Our analysis revealed significant regional differences in QnR eco resistance and detection rates across China. Along the Hu Huanyong Line, resistance rates varied markedly: 49.35 in the northwest, 54.40 on the line, and 52.30 in the southeast (P = 0.001). Detection rates also showed significant geographical variation, with notable differences between regions (P < 0.001). Climate types influenced these rates, with significant variability observed across different climates (P < 0.001). Our predictive model for resistance rates, integrating climate and healthcare factors, explained 64.1% of the variance (adjusted R-squared = 0.641). For detection rates, the model accounted for 19.2% of the variance, highlighting the impact of environmental and healthcare influences. The study found higher resistance rates in warmer, monsoon climates and areas with more public health facilities, but lower rates in cooler, mountainous, or continental climates with more rainfall. This highlights the strong impact of climate on antibiotic resistance. Meanwhile, the predictive model effectively forecasts these resistance rates using China's diverse climate data. This is crucial for public health strategies and helps policymakers and healthcare practitioners tailor their approaches to antibiotic resistance based on local environmental conditions. These insights emphasize the importance of considering regional climates in managing antibiotic resistance. [Display omitted] • E. coli 's resistance rates change with climate and Hu line. • Region-specific health strategies against antibiotic resistance is needed. • The model has a higher explanatory power. [ABSTRACT FROM AUTHOR]

• End-to-End solution of memory reliability in data center environment combing ML-models and engineering tools. • Memory error pattern modeled as images to apply CNN models. • Providing a visualization tool for memory errors on cell level. • Providing an MLOps framework for operating the CNN model in production. • Concept for self-supervised data labeling. With the trend towards multi-socket server systems, the demand for random access memory (RAM) per server increased. The consequence are more DIMM sockets per server. Since every dual in-line memory module (DIMM), which comprises a series of dynamic random-access memory integrated circuits, has a probability of failure, RAM issues became a dominant failure pattern for servers. The concept introduced in this work contributes to improving the reliability of data centers by avoiding RAM failures and mitigating their impact. For this purpose, an ML-driven framework is provided to estimate the probability of memory failure for each RAM module. The ML framework is based on structural information between correctable (CE) and uncorrectable errors (UE). In a common memory scenario, a corrupted bit within a module can be restored by redundancy using an error correction code (ECC), resulting in a CE. However, if there is more than one corrupted bit within a group of bits covered by the ECC, the information cannot be restored, resulting in a UE. Consequently, the related task requesting the memory content, and the corresponding service may crash. There is evidence that UEs have a CE history and structural relation between the CEs. However, for the case of UEs without a CE history or of a false decision of the ML framework, we extend the total framework by engineering measures to mitigate the impact of a UE by avoiding kernel panic and using backups. The engineering measures use a mapping between physical and logical memory addresses. [ABSTRACT FROM AUTHOR]

• Model: Proposed a new model to handle different SLAs of workloads and guarantee the worst-case delay of delay-tolerant workload. • Algorithm: Designed an O (1) time online control algorithm OSDG for each slot to achieve the long term optimization objective. • Analysis and Experiment: Proved our algorithm achieves an explicit [ O (1 / V) , O (V) ] cost-delay tradeoff and validated its performance. More and more cloud data centers are turning to leverage on-site renewable energy to reduce power cost for sustainable development. But how to effectively coordinate the intermittent renewable energy with workload remains to be a great challenge. This paper investigates the problem of workload scheduling for power cost minimization under the constraints of different Service Level Agreements (SLAs) of delay tolerant workload and delay sensitive workload for green data centers in a smart grid. Different from the existing studies, we take into consideration of the impact of zero price in the smart grid and the cost of on-site renewable energy. To handle the randomness of workload, electricity price and renewable energy availability, we first formulate the problem as a constrained stochastic problem. Then we propose an efficient online control algorithm named ODGWS (Online Delay-Guaranteed Workload Scheduling) which makes online scheduling decisions achieve a bounded guarantee from the worst scheduling delay for delay tolerant workload. Compared with the existing solutions, our ODGWS decomposes the problem into that of solving a simple optimization problem within each time slot in O (1) time without needing any future information. The rigorous theoretical analysis demonstrates that our algorithm achieves a [ O (1 V) , O (V) ] cost-delay tradeoff, where V is a balance parameter between the cost optimality and service quality. Extensive simulations based on real-world traces are done to evaluate the performance of our algorithm. The results show that ODGWS saves about 5% average power cost compared with the baseline algorithms. [ABSTRACT FROM AUTHOR]

Modern virtualized data centers often rely on virtual machine (VM) migrations to consolidate workload on a single machine for energy saving. But VM migrations have many drawbacks, including performance degradation, service disruption etc. Hence, many approaches have been proposed to minimize the overhead when migrations occur. In contrast, this work aims to proactively avoid migrations from happening in the first place. We have proposed a novel consolidation aware scheduling algorithm to minimize the number of migrations for batch processing systems by taking advantage of the prior knowledge of consolidation strategy and job information. We show the problem can be formulated as an integer linear programming (ILP) problem, and an effective heuristic solution can be found by a genetic algorithm. Both real and synthetic workload traces were used to evaluate our methods. Experimental results showed that, after comparing with two popular job scheduling algorithms, our approach has reduced the number of migrations by more than 25%. • Proactively avoid VM migration in virtualized data centers. • A co-design between consolidation and scheduling decisions. • Construct an ILP formulation to find an optimal solution. • Apply genetic algorithm to find a heuristic solution. • Using real world workload traces from parallel computing cluster for experimentation. [ABSTRACT FROM AUTHOR]

Communication in data centers often involves many parallel flows that all share the same performance goal (e.g. to minimize the average completion time). A useful abstraction, coflow , is proposed to express the communication requirements of prevalent data parallel paradigms such as MapReduce and Spark. The multiple coflow routing and scheduling problem makes it challenging to derive a good theoretical performance ratio, as coexisting coflows may compete for the same network resources such as link bandwidths. In this paper, we focus on the coflow problem in one popular data center infrastructure: the Leaf-Spine topology. We first formulate the problem and study the path selection issue on this two-tier structure. In order to minimize the average coflow completion time (CCT), we propose the Multi-hop Coflow Routing and Scheduling strategy (MCRS) and prove that our method has a reasonably good competitive ratio. Extensive experiments and large-scale simulations show that MCRS outperforms the state-of-the-art heuristic schemes under the Leaf-Spine topology. [ABSTRACT FROM AUTHOR]

Coflow is a network abstraction used to represent communication patterns in data centers. The coflow scheduling problem encountered in large data centers is a challenging NP -hard problem. Many previous studies on coflow scheduling mainly focus on the single-core model. However, with the growth of data centers, this single-core model is no longer sufficient. This paper addresses the coflow scheduling problem within heterogeneous parallel networks, which feature an architecture consisting of multiple network cores running in parallel. In this paper, two polynomial-time approximation algorithms are developed for the flow-level scheduling problem and the coflow-level scheduling problem in heterogeneous parallel networks, respectively. For the flow-level scheduling problem, the proposed algorithm achieves an approximation ratio of O (log ⁡ m / log ⁡ log ⁡ m) when all coflows are released at arbitrary times, where m represents the number of network cores. On the other hand, in the coflow-level scheduling problem, the proposed algorithm achieves an approximation ratio of O (m (log ⁡ m / log ⁡ log ⁡ m) 2) when all coflows are released at arbitrary times. Moreover, we propose a heuristic algorithm for the flow-level scheduling problem. Simulation results using synthetic traffic traces validate the performance of our algorithms and show improvements over the previous algorithm. • Algorithm achieves O (log ⁡ m / log ⁡ log ⁡ m) approximation ratio for flow-level scheduling, where m is network cores. • For coflow-level scheduling, algorithm achieves O (m (log ⁡ m / log ⁡ log ⁡ m) 2) approximation ratio. • Moreover, this paper proposes a heuristic algorithm for the flow-level scheduling problem. [ABSTRACT FROM AUTHOR]

As an essential infrastructure to support cloud computing services, data center networks (DCNs) are primarily used for transmitting and storing data. BCube is a large-scale DCN architecture that can effectively handle the massive data generated by network end devices due to the explosive growth of the Internet. However, as server failures in DCNs become more frequent, ensuring reliable data communication in BCube is critical. In this paper, we first establish a conditional local safe model in BCube, which divides the fault-free nodes in sub-BCube by adding some restrictions, and effectively avoids communication obstacles that may arise from faulty nodes. Then, based on the proposed model, a fault-tolerant unicast path algorithm is designed, which can construct a reliable data transmission path in the local safe sub-BCube of BCube and can tolerate more faulty nodes than existing algorithms. Finally, we conduct simulation experiments to verify that our algorithm can construct the shortest path with a high probability and achieve almost complete success in data transmission when the number of faulty nodes does not exceed half of the total nodes in BCube. • We propose a conditional local safe model in the data center network BCube. • We prove the existence of fault-free paths with local safe information in BCube. • We construct a fault-tolerant unicast algorithm in BCube based on the above model. [ABSTRACT FROM AUTHOR]

Fog computing is an emerging paradigm that brings computing capabilities closer to distributed IoT devices, which provides networking services between end devices and traditional cloud data centers. One important mission is to further reduce the monetary cost of fog resources while meeting the ever-growing demands of multiple users. In this paper, we focus on minimizing the total cost for multiple mobile users to provide an efficient resource provisioning scheme in fog computing. The total cost includes two aspects: the replication cost and the transmission cost. We consider three cases for the resource provision problem by focusing on different cost models. First, one simple case where users can only upload one replication is discussed, and an optimal solution is proposed by converting the original problem into a bipartite graph matching. Then we consider a more complicated case in which each user can upload multiple replications on fog nodes in the resource provisioning. Specifically, two models are discussed: the 0–1 transmission cost model and the different transmission cost model. For the 0–1 transmission cost model, each user can upload multiple replications with a constant transmission cost, and one optimal greedy solution is proposed. For the different transmission cost model, the transmission cost is related to the distance of each pair of fog nodes. This problem is proven to be NP-hard. We first propose a non-adaptive algorithm which is proved to be bounded by 2 3 W + 1 3 O P T. Another 3 + ϵ -approximation algorithm is proposed based on local search, which has better performance with higher complexity. Extensive simulations also prove the efficiency of our schemes. • We consider the resource provision problem for users with cost minimization and model the identified problem by considering two types of cost: replication cost and transmission cost. We first discuss a simple case, and one optimal solution is proposed. We then consider a more complicated case in which each user can upload multiple replications on fog nodes in the provisioning process. Two models are discussed: the 0–1 transmission cost model and the different transmission cost model. • We propose an optimal solution for the 0–1 transmission cost model. Then we prove that the problem under the different transmission cost model is NP-hard. We first propose a non-adaptive algorithm which is proved to be bounded by 2 3 W + 1 3 O P T. Another 3 + ϵ -approximation algorithm is proposed based on local search, which has better performance with higher complexity. • Various simulations were conducted to compare our joint optimization methods with several state-of-the-art ones based on two real datasets: the Mobike dataset and the Microsoft GPS trajectory dataset. The results are shown from different perspectives to provide conclusions. [ABSTRACT FROM AUTHOR]

Elephant flows (elephants) refer to the sequences of packets that contribute only 10% of the total volume but consume over 90% of the network bandwidth. They often cause network congestion and should be efficiently managed. Present cloud data centers often involve host- and switch-based approaches to detect and schedule elephants, but suffer (1) each host and switch in the network needs to be customized, and (2) dynamic models and advanced policies are difficult to be applied. Software Defined Cloud (SDC) addresses these issues by enabling controller-based approaches. With the aid of Machine Learning (ML) technologies, SDC can achieve learning-based models, flexible deployment, and early detection and schedule of elephants for the optimization of network performance and energy usage in a dynamic and intelligent manner. On this purpose, this article emphases the significance of models describing elephants, surveys the mechanisms that may apply to model, detect, and schedule elephants for SDC to optimize the network performance and energy usage. To the best of our knowledge, this work is the first effort that reviews the techniques in all these related subtopics simultaneously in the context of energy cloud. • This study surveys how to optimize the network performance and energy usage for SDC. • Review the techniques in related subtopics simultaneously in the context of energy cloud and machine learning. • A survey mechanism is developed based on the modeling, detecting, and scheduling elephant flows of SDC. [ABSTRACT FROM AUTHOR]

Large-scale data-intensive applications provide services to users by routing service requests to geographically distributed data centers interconnected by Internet links. In order to achieve good reliability and data access latency performance, cloud service providers often simultaneously place multiple copies of the data in different data centers. The network communication required for updating the multiple data copies incurs an operational cost. At the same time, the penalty incurred by the Service Level Agreement (SLA) violation for data access from the data centers also imposes an operational cost on the service providers. In this paper, we tackle the problem of data placement in distributed data centers with the aim to minimize the operational cost incurred by delay SLA violation penalty and inter-data center network communication, assuming each data has K data replicas. We propose a K-level Cluster-based Data Placement algorithm (K-CDP) for the problem. The algorithm solves the linear programming relaxation and dual programming problems corresponding to the problem of minimizing SLA violation penalty cost caused by placing a replica of each data in a data center. Based on the obtained solutions, the algorithm clusters the data so that the data with similar placeable data centers form a data cluster. For the data in each cluster, the algorithm selects K data centers to minimize the operational cost. We prove that algorithm K-CDP is 2-approximation to the data placement problem. Our simulation results demonstrate that the proposed algorithm can effectively reduce the penalty cost incurred by delay SLA violation, the network communication cost, and the operational cost of data centers. • We tackle the data placement problem to minimize SLA violation and network communication cost. • We propose an algorithm which is proved to be 2-approximation to the data placement problem. • The simulation results show the proposed algorithm is effective and promising. [ABSTRACT FROM AUTHOR]

This work presents Speculative Transaction Replication (STR), a protocol that exploits transparent speculation techniques to enhance performance of geo-distributed, partially replicated transactional data stores. In addition, we define a new consistency model, Speculative Snapshot Isolation (SPSI), a variant of Snapshot Isolation (SI), which shelters applications from the subtle anomalies that arise when using speculative transaction processing techniques. STR provides a form of speculation that is fully transparent for programmers (it does not expose the effects of misspeculations to clients). Since the speculation techniques employed by STR satisfy SPSI, they can be leveraged by application programs in a transparent way, without requiring any source-code modification to applications designed to operate using SI. STR combines two key techniques: speculative reads, which allow transactions to observe pre-committed versions, which can reduce the 'effective duration' of pre-commit locks and enhance throughput; Precise Clocks, a novel timestamping mechanism that uses per-item timestamps with physical clocks, which together greatly enhance the probability of successful speculation. We assess STR's performance on up to nine geo-distributed Amazon EC2 data centers, using both synthetic benchmarks as well as realistic benchmarks (TPC-C and RUBiS). Our evaluation shows that STR achieves throughput gains up to 11 × and latency reduction up to 10 × (with respect to non-speculative systems that ensure SI) in workloads characterized by low inter-data center contention. Furthermore, thanks to a self-tuning mechanism that dynamically and transparently enables and disables speculation, STR offers robust performance even when faced with unfavorable workloads that suffer from high misspeculation rates. • We propose Speculative Snapshot Isolation, a consistency model that provides clear and rigorous guarantees on the atomicity and isolation of the snapshots observed and produced by speculative transactions. • We propose STR, a speculative transactional protocol for partially replicated geo-distributed data stores. STR is a full y decentralized protocol that ensures (SPSI-)safe speculations. • The evaluation study on nine AWS data centers shows that the use of speculation yields up to 11x throughput gain. [ABSTRACT FROM AUTHOR]

In the age of the information explosion, the energy demand for cloud data centers has increased markedly; hence, reducing the energy consumption of cloud data centers is essential. Dynamic virtual machine VM consolidation, as one of the effective methods for reducing energy energy consumption is extensively employed in large cloud data centers. It achieves the energy reductions by concentrating the workload of active hosts and switching idle hosts into low-power state; moreover, it improves the resource utilization of cloud data centers. However, the quality of service (QoS) guarantee is fundamental for maintaining dependable services between cloud providers and their customers in the cloud environment. Therefore, reducing the power costs while preserving the QoS guarantee are considered as the two main goals of this study. To efficiently address this problem, the proposed VM consolidation approach considers the current and future utilization of resources through the host overload detection (UP-POD) and host underload detection (UP-PUD). The future utilization of resources is accurately predicted using a Gray-Markov-based model. In the experiment, the proposed approach is applied for real-world workload traces in CloudSim and were compared with the existing benchmark algorithms. Simulation results show that the proposed approaches significantly reduce the number of VM migrations and energy consumption while maintaining the QoS guarantee. • The proposed Virtual Machine consolidation approach reduces the power costs while preserving the quality of service (QoS) guarantee. • Utilization of resources is accurately predicted using a new model. • In the experiment, the proposed approach is applied for real-world workload traces in CloudSim and were compared with the existing benchmark algorithms. • Simulation results show that the proposed approaches significantly reduce the number of VM migrations and energy consumption while maintaining the QoS guarantee. [ABSTRACT FROM AUTHOR]

Live migration of virtual machines (VMs) has become an extremely powerful tool for cloud data center management and provides significant benefits of seamless VM mobility among physical hosts within a data center or across multiple data centers without interrupting the running service. However, with all the enhanced techniques that ensure a smooth and flexible migration, the down-time of any VM during a live migration could still be in a range of few milliseconds to seconds. But many time-sensitive applications and services cannot afford this extended down-time, and their clocks must be perfectly synchronized to ensure no loss of events or information. In such a virtualized environment, clock synchronization with minute precision and error boundedness are one of the most complex and tedious tasks for system performance. In this paper, we propose enhanced DTP and wireless PTP based clock synchronization algorithms to achieve high precision at intra and inter-cloud data center networks. We thoroughly analyze the performance of the proposed algorithms using different clock measurements. Through simulation and real-time experiments, we also show the effect of various performance parameters on the data center networking architectures. • Clock synchronization problem for time-sensitive applications and services. • Formulated various measures for efficient synchronization at different DC levels. • Present enhanced DTP and wireless PTP based algorithms for live VM migration. • Evaluate optimal precision with different loss measurements. [ABSTRACT FROM AUTHOR]

Stellar datacenter networks are a recent generic construction designed to transform a base-graph into a dual-port, server-centric datacenter network. We prove that the S -bisection width of any stellar datacenter network can be obtained from the solution of isoperimetric problems on the base-graph, provided that the base-graph is regular. We extend previous research on the stellar datacenter networks GQ⁎, instantiated with generalized hypercubes, and show that with respect to S -bisection width, GQ⁎ performs well in comparison with the dual-port datacenter network FiConn. Our work develops a strong combinatorial link between graph bisection width and throughput metrics for stellar datacenter networks. [ABSTRACT FROM AUTHOR]

With the rise of cloud service providers and the continuous virtualization of data centers, data center networks are also developing rapidly. As data centers become more and more complex, the demand for security increases dramatically. This paper discusses the privacy inherent in data centers. However, there is no general solution to the privacy problem in data centers due to the device heterogeneity. In this paper, we proposed a local differential privacy-based classification algorithm for data centers. In data mining of data centers, the differential privacy protection mechanism is added to deal with Laplace noise of sensitive information in the pattern mining process. We designed a method to quantify the quality of privacy protection through strict mathematical proof. Experiments demonstrated that the differential privacy-based classification algorithm proposed in this paper has higher iteration efficiency, better security and feasible accuracy. On the premise of ensuring availability, the algorithm has reliable privacy protection characteristics and excellent timeliness. • First, we proposed a local differential privacy-based classification algorithm for data centers. In data mining of data centers, the differential privacy protection mechanism is added to deal with Laplace noise of sensitive information in the pattern mining process. • Second, We proposed a classification algorithm satisfies local differential privacy for cloud computing supported data centers. It can guarantee the validity and reliability of privacy protection and achieves privacy preserving by adding Laplacian mechanism. Experimental results demonstrated that the proposed algorithm has excellent reliability and efficiency. [ABSTRACT FROM AUTHOR]

Cloud computing offers attractive features for both service providers and customers. Users benefit from the pay-as-you-go model by saving expenditures and service providers are deploying their services to cloud data centers to reduce their maintenance efforts. However, due to the fast growth of cloud data centers, the energy consumed by the data centers can lead to a huge amount of carbon emission with environmental impacts, and the carbon intensity of different locations are varied among different power plants according to the sources of energy. Thus, in this paper, to address the carbon emission problem of data centers, we consider shifting the workloads among multi-cloud located in different time zones. We also formulate the energy usage and carbon emission of data centers and model the solar power corresponding to the locations. This helps to reduce the usage of brown energy and maximize the utilization of renewable energy at different locations. We propose an approach for managing carbon footprint and renewable energy for multiple data centers at California, Virginia, and Dublin, which are in different time zones. The results show that our proposed approaches that apply workload shifting can reduce around 40% carbon emission in comparison to the baseline while ensuring the average response time of user requests. • Model considering power consumption, renewable energy and carbon footprint. • Algorithm for managing the renewable energy and carbon footprint for data centers. • Evaluations on the impacts of requests executed in different time zones. [ABSTRACT FROM AUTHOR]

Data centers form the core of current information and commercial enterprise. At current scale, any improvement in data center resource utilization leads to substantial savings. We focus on the problem of scheduling jobs in distributed execution environments to improve resource utilization. Cluster schedulers like YARN and Mesos base their scheduling decisions on resource requirements provided by end users. It is hard for end-users to predict the exact amount of resources required for a task/job, especially since resource utilization can vary significantly over time and across tasks. In practice, users make highly conservative estimates of peak utilization across all tasks of a job to ensure job completion, leading to resource fragmentation and severe under utilization in production clusters. We present UBIS, a utilization-aware approach to cluster scheduling, to address resource fragmentation, and to improve cluster utilization and job throughput. UBIS considers actual usage of running tasks and schedules opportunistic work on under-utilized nodes. UBIS monitors resource usage on these nodes and preempts opportunistic containers in the event this over-subscription becomes untenable. In doing so, UBIS effectively utilizes wasted resources, while minimizing adverse effects on regularly scheduled tasks. Our implementation of UBIS on YARN demonstrates improvements of up to 30% in makespan for a representative workload and 25% in individual job durations. • The paper presents a scheduling technique, called UBIS for cloud environments with mixed jobs – some with guaranteed service level requirements (SLAs), and others with flexible requirements. • UBIS utilizes slack in runtime utilization of resources by tasks with guaranteed service level requirements to schedule opportunistic tasks around them. • These opportunistic tasks are terminated if SLAs are likely to be violated, as inferred from dynamic state of the nodes. • By utilizing slack, UBIS significantly increases resource utilization and throughput, while ensuring fairness in highly dynamic environments. • Comprehensive benchmarking of UBIS in cloud deployments with realistic workloads demonstrates significant increase in throughput with associated operational cost reduction, low overhead of monitoring and scheduling, and excellent scalability. [ABSTRACT FROM AUTHOR]

Due to the application in reliable communication, reliable broadcasting, secure message distribution, etc., node/edge-independent spanning trees (ISTs) have attracted much attention in the past twenty years. However, node/edge conjecture is still open for networks with node/edge-connectivity ≥ 5. So far, results have been obtained on a lot of special networks, but only a few results are reported on the line graphs of them. Hypercubes play important roles in parallel computing systems, and the line graphs of which have been recently adopted for the architectures of data center networks. Since the line graph of n -dimensional hypercube Q n , L (Q n) , is (2 n − 2) -regular, whether there exist 2 n − 2 node-ISTs rooted at any node on L (Q n) is an open question. In this paper, we focus on the problem of constructing node-ISTs on L (Q n). Firstly, we point out that L (Q n) can be partitioned into 2 n − 1 complete graphs. Then, based on the complete graphs and n − 1 node-ISTs rooted at 0 on Q n − 1 0 , we obtain an "independent forest" containing 2 n − 2 trees on L (Q n). Furthermore, we present an O (N) time algorithm to construct 2 n − 2 node-ISTs rooted at node [0, 2 n − 1 ] isomorphic to each other on L (Q n) based on the independent forest, where N = n × 2 n − 1 is the number of nodes on L (Q n). In addition, we point out that the 2 n − 2 node-ISTs on L (Q n) is a new method to prove the node/edge-connectivity and the upper bound of (2 n − 2) -node/edge-wide-diameter of L (Q n). • We propose the definition of independent forest and use it to solve the node-independent spanning trees (ISTs) problem on L (Q n). • We design an O (N) time algorithm to construct 2 n − 2 node-ISTs rooted at node [0, 2 n − 1 ] in L (Q n) , where N = n × 2 n − 1 is the number of nodes in L (Q n). • The height of the node-ISTs is n + 2 and the trees are isomorphic to each other. • The 2 n − 2 node-ISTs in L (Q n) is a new method to prove the node/edge connectivity and the upper bound of (2 n − 2) -node/edge-wide-diameter of L (Q n). [ABSTRACT FROM AUTHOR]

The computing power of supercomputers and data centers has noticeably grown during the last decades at the cost of an ever increasing energy demand. The need for energy (and power) of these facilities has finally limited the evolution of high performance computing, making that many researchers are concerned not only about performance but also about energy efficiency. However, despite the many concerns about energy consumption, the search for computing power continues. In this regard, the research on exascale systems, able to deliver 1018 floating point operations per second, has reached a widely consensus that these systems should operate within a maximum power budget of 20 megawatts. Many efficiency improvements are necessary for achieving this goal. One of these improvements is the usage of ARM low-power processors, as the Mont-Blanc project proposes. In this paper we analyze the combined use of ARM processors with the rCUDA remote GPU virtualization middleware as a way to improve efficiency even more. Results show that it is possible to speed up applications by almost 8x while reducing energy consumption up to 35% when rCUDA is used to access high-end GPUs. These improvements are achieved while maintaining a feasible average power consumption level for future exascale systems. • Suitability of rCUDA remote GPU virtualization middleware for future exascale. • Performance, power and energy consumption measurements. • Almost 8x application speed up while reducing energy consumption up to 35% • Maintaining a feasible average power consumption level for future exascale systems. [ABSTRACT FROM AUTHOR]

Data center performance is affected by three main factors; bandwidth, latency, and reliability of intra-data center interconnection network. Bandwidth and latency are definitely improved by adopting optical technology for intra-data center communication, but fault tolerance of the corresponding optical networks has been raised less. Recently, we introduced two Torus-based, all-optical, and non-blocking networks, i.e. O-TF and O-FTF, addressing reliability of optical networks, and now, in this paper, to address the scalability problem, we propose a novel Optical Clos-based architecture which reduces minimum number of required wavelength channels, as well as, the switch size in each node. Moreover, we examine three different types of redundancies; including (a) wavelength (adopted in O-TF network), (b) hardware (adopted in O-FTF network), and (c) path (adopted Optical Clos)) redundancies, in terms of throughput drop under failures, implementation cost, optical power loss, electrical power, latency, topology, and wiring complexity. In addition to simulating networks utilizing generated traffic based on Microsoft data, realistic traffic has also been considered. Our simulation results confirm that in addition to improving power consumption and latency, O-TF, O-FTF, and O-Clos architectures lead to throughput increment up to 80%, 90%, and 70%, respectively, comparing to WaveCube at the presence of network failure, as result of adopting our proposed fault-tolerant method, named as RAOP. [ABSTRACT FROM AUTHOR]

To promote the intelligent and accurate management of river basins, especially large basins which involve many catchments, it is highly required to develop a useful platform to effectively coordinate arithmetic resources and data, and simultaneously help to make decisions based on the real-time calculation. In this study, a multi-centre cloud platform architecture called 3L4C was constructed, which includes a Cloud-edge-terminal Layer (3L), data centre, model centre, control centre, and customer-service centre (4C). Data fusion technology and an air-land-water coupled model were constructed. Based on HTML5, JavaScript, and Java, an integrated water environment management platform was created and applied to the Three Gorges Reservoir Basin, China. The platform was tested and successfully used for automatic water quality prediction, water environment pollution analysis and control, early warning of abnormal water quality, and emergency water pollution incident evaluation. This platform quickly and accurately forecasts and perfectly displays past, present and future state of the water environment, and offers beneficial support for management decisions in various water environment departments. • A cloud platform architecture for environmental management decision-making. • Three Gorges Reservoir Basin Water Environment Management Platform. • Integrate water environment data from multiple sources. [ABSTRACT FROM AUTHOR]

• An optical band-pass filter-assisted offset DSB direct-detection receiver is proposed for low-cost data center interconnect applications. • The proposed scheme uses an OBPF to filter out the optical carrier as the local oscillator for the receiver. As such, signal-signal beat interference can be eliminated. • Compared with the previous offset DSB CADD scheme, the optimal CSPR of the proposed scheme can be reduced to less than 0 dB, and the OSNR sensitivity can be significantly improved. • The effectiveness of the proposed offset DSB self-coherent detection is fully verified by simulation. The numerical results show the potential of proposed scheme. The offset double-sideband (DSB) scheme has attracted attention because it can achieve field recovery at low carrier-to-signal power ratio (CSPR) and improve the spectral efficiency (SE) of self-coherent detection. In this paper, an offset DSB filter-assisted direct detection (FADD) scheme is proposed and studied for the first time, which uses an optical bandpass filter (OBPF) to filter out the optical carrier as the local oscillator for the receiver. As such, signal-signal beat interference (SSBI) can be eliminated. Since the offset DSB signal has a large frequency gap, the requirement for the bandwidth of the OBPF is not high. The performance of the proposed offset DSB FADD scheme is evaluated for 16-QAM signals. It is comprehensively compared with offset DSB carrier-assisted differential detection (CADD) scheme in terms of OSNR sensitivity, optimal CSPR, and receiver power sensitivity. In addition, the effect of frequency gap compression is significant as SSBI is completely eliminated from the system, thereby utilizing as many low-frequency resources as possible and improving SE. Finally, the proposed offset DSB FADD scheme is resilient to the wavelength offset up to several GHz between the transmitter laser and the center wavelength of the OBPF. [ABSTRACT FROM AUTHOR]

Meat consumption and convenience food are both located at the heart of contemporary, industrialized, unhealthy and unsustainable food systems. In this article, we study the intersections between convenience food and 'meatification' of diets, focusing on the 'pølse'—an umbrella term including both hotdogs and a range of sausages—as the epitome of convenience food in Norway. We explore how the pølse is embedded in Norwegian food practices, and why it is considered convenient in different contexts. In doing so, we seek to explain how pølse eating is co-shaped by socio-material scripting processes that further entrench meat in food practices and complicate meat-reduction efforts. The analysis is based on 52 in-depth household interviews and autophotography in four geographical contexts in Norway, in addition to 22 park interviews and survey data centering on household food and meat practices. We use a theoretical apparatus combining social practice theory, foodscapes and socio-material scripts to analyse the conveniencization of pølse. The articles demonstrates how meat consumption and convenience food become entwined in specific social practices, and how conveniencization intersects with practices of care, notions of class, social expectations and normativity. Moreover, we show that despite the range of plant-based 'pølse' substitutes on offer, meat-eaters remain skeptical to its taste, and substitutes rather seem to offer a way into established social occasions for non-meat eaters than a way out of meat eating. [ABSTRACT FROM AUTHOR]

Data center is a very important infrastructure to support the development of information technology, and its development and increment are very remarkable. However, with the rapid and large-scale development of data centers, the problem of energy consumption turns to be also very prominent. Under the background of global carbon peak and carbon neutrality, developing green and low-carbon data centers has become an inevitable trend. This paper reviews and analyzes the policies and their roles in promoting China's green development of data centers in the past 10 years, summarizes the current situation of the implementation of green data center projects in China and gives the changes of PUE limits of data centers under the policy constraints. Application of green technologies is an important measure for energy-saving and low-carbon development of data centers, so encouraging innovation and application of green technologies in data center is also a priority task in relevant policies. This paper points out the green and low-carbon technology system of data centers, further summarizes energy-saving and carbon-reducing technologies in IT equipment, cooling system, power supply and distribution system, lighting, intelligent operation and maintenance, and provides an outlook on the future green development of data centers. • China's green data center policy in recent 10 years has played an important role in promoting the development of green data centers. • Summarizes the status of green data centers under policy guidance. • Provides energy saving and carbon reduction technology path for data centers. • Prospects for the future development of green data centers. [ABSTRACT FROM AUTHOR]

• A low-power solution for short-distance high-speed data center interconnects. • Self-homodyne (SH) systems provide a line-width tolerant and LO-less receiver. • Applicability of a proposed cascaded equalizer is verified by analysis and results. • A self-homodyne QPSK system with cascaded equalizer is experimentally demonstrated. • Analog implementation of such equalizer results in a low-power lowcost receiver. Self-homodyne fiber optic links appear to be ideal for high-capacity low-power data center interconnects. In this work, a polarization multiplexed carrier based Self-Homodyne Quadrature Phase Shift Keying (SH-QPSK) system has been experimentally demonstrated with a cascaded equalizer structure. The proposed cascade combination of a Constant Modulus Algorithm (CMA) based equalizer with a Decision Feedback Equalizer (DFE) improves the performance by virtue of the difference of the Mean Squared Error (MSE) of the two equalizers. The reduction in the MSE for a CMA equalizer, when an additional DFE stage is used, is confirmed by theoretical framework and simulation results. Simulation results also validate that the difference in the performance measurement for the standalone CMA and the cascaded equalizer is more in the case of such SH system as compared to the conventional coherent system. The 25 Gbaud SH-QPSK system results in a Bit Error Rate (BER) of 4.14 × 10 - 6 for 80 km Standard Single Mode Fiber (SSMF) with the CMA-DFE equalizer. The CMA-DFE equalizer exhibits one to two orders of magnitude improvement in the BER for various lengths of SSMF channel, as compared to the BER obtained from a commonly used standalone CMA equalizer. [ABSTRACT FROM AUTHOR]

• Orchestrated deployment of slices over a data centre with optical DCN is presented. • The optical data plane allows for high bandwidth and low latency communications. • The resource orchestration approach increases the number of accepted slices. The emergence of new paradigms and services is pushing the limits of nowadays cloud infrastructures. It is a fact that current solutions lack in the flexibility and configurability to adapt to heterogeneous requirements coming from the applications/services to be supported over them. This results most of the time in severe underutilization of the underlying physical substrate. In light of this, newer approaches on resource provisioning and infrastructure management are needed. The Infrastructure as a Service (IaaS) paradigm is proposed as a solution to overcome these limitations. Thanks to IaaS, the physical infrastructure is partitioned onto virtual slices, encompassing heterogeneous resources (e.g. network, computing). Such a concept is expected to be harnessed by future data center (DC) infrastructures in order to cope efficiently with multi-tenancy as well as heterogeneous application requirements. However, current DC networks (DCNs) impose sever limitations onto traffic handling to fully exploit this vision. In light of this, optical technologies are seen as prime candidates for realizing the high performance network fabrics that future DC architectures will need. Under such an umbrella, it becomes primordial to develop specific provisioning solutions that account for the particularities of the optical medium, while providing the means to efficiently slice the DC infrastructure. With all of these in mind, in this article we present a solution for orchestrating and controlling virtual slices in a DC scenario with optical intra-DCN, with the scope of optimizing the underlying physical infrastructure utilization. The benefits of the presented solution are demonstrated against legacy architectures through exhaustive experiments and simulations. [ABSTRACT FROM AUTHOR]

Abstract The development of the cloud and data centers alike has caused users to increasingly share storage systems, and storage systems have adopted SSDs to improve performance. The number of users is growing sharply and causing resource competition, so the throughput allocation of a hybrid storage system has an important effect on users' performance. A hybrid storage architect faces a challenge: sharing I/O throughput resources between SSDs and HDDs fairly with high resource utilization. Drawing on economic game-theory, many methods based on Leontief preference have been proposed to determine fair throughput allocation in a hybrid storage system with having the properties of Sharing Incentive (SI), Envy Freeness (EF) and Pareto Optimality (PO). However, users' workload characteristics vary because of the adoption of SSDs, so there is heterogeneity among the storage media. The heterogeneity of storage media and the diversity of user workloads together create more constraints on fairness, which imposes restrictions to improve the resource utilization. To improve the resource utilization, some researchers have proposed relaxing the limitation of fairness properties by adjusting users' allocations. However, when adjusting the allocation between users, some users' allocations are reduced but others' are improved, with the goal of increasing the resource utilization of the whole system, this will pose a problem because of the Leontief preference restriction: some users' performance is degraded, while other users' performance is improved. In this paper, we first propose one variant of the Cobb–Douglas preference based on users' workloads and allocations, and then present a throughput allocation model based on this variant named TACD that can capture the effects of diminishing returns and substitutability in throughput allocation for hybrid storage. By making use of these effects, TACD can improve some users' performance while providing the same performance for other users and achieving higher resource utilization in the whole system, and it can provide a fair allocation while enjoying SI, EF and PO properties. The variant of the Cobb–Douglas preference proposed in this paper has taken consideration of users' workloads, so TACD can conduct the throughput allocation based on these workloads for meeting the performance requirements of users reasonably. Extensive experiments are conducted to prove the effectiveness of TACD. The simulation results show that TACD can fit the throughput allocation of hybrid storage system very well. The Linux results show that compared with the allocation method based on Leontief preference, TACD can provide a higher resource utilization and performance for users throughout the system, it can improve some users' performance when maintaining comparable performance for other users, in contrast to the Leontief method based on weaker fairness under the same resource utilization, and its allocation result enjoys SI, EF and PO properties. Highlights • The throughput allocation method based on the variant of Cobb–Douglas preference (TACD). • The variant of Cobb–Douglas preference based on users' workloads, which can conduct a reasonable allocation for meeting users' performance. • TACD can capture the effects of diminishing returns and substitutability for improving the resource utilization with enjoying fairness properties, and its allocation can improve some users' performance without degrading the performance of the remaining users. [ABSTRACT FROM AUTHOR]

Abstract In a data-rich digital world, our hand-held resource-constrained mobile devices are restricted to performing small-to-medium-level computation processes and are incapable of performing high-computation processes. Computation offloading is a suitable solution for overcoming this shortcoming. Until recently, we have perceived cloud computing as an appropriate computation-offloading platform for mobile devices. However, cloud data centers, being far-end networks for mobile devices, increase the latency or network delay, which in turn affects the performance of real-time mobile Internet-of-Things applications. Hence, for critical real-time applications, a near-end network approach of computation offloading is required. Furthermore, the major hurdles for geographically distributed mobile devices are mobility and heterogeneity in the process of computation offloading. To overcome these challenges, the use of a deep-learning-based response-time-prediction framework is proposed in this paper to determine whether to offload in the nearby fog/edge node or neighbor fog/edge node, or cloud node. Furthermore, a restricted Boltzmann machines learning is applied to tackle the randomness in the availability of resources. We simulate the proposed model in MATLAB while considering the mobility and fluctuating resource demands of the end users. Implementing our deep-learning-based response-time-prediction framework improves the performance of the computation offloading because it facilitates a prompt selection of the offloading location. Highlights • An efficient computation offloading model for edge/cloud is proposed. • A novel machine learning model is used for computation offloading. • Our method significantly improves the performance of the computation offloading. [ABSTRACT FROM AUTHOR]

Abstract Live Virtual Machine (VM) migration within a data center is an important technology for cloud management, and has brought many benefits to both cloud providers and users. With the development of cloud computing, across-data-center VM migration is also desired. Normally, there is no shared storage system between data centers, hence the storage data (disk image) of a VM will be migrated to the destination data center as well. However, the slow network speed of the Internet and the comparatively large size of VM disk image make VM storage data migration become a bottleneck for live VM migration across data centers. In this paper, based on a detailed analysis of VM deployment models and the nature of VM image data, we design and implement a new migration system, called CBase. The key concept of CBase is a newly introduced central base image repository for reliable and efficient data sharing between VMs and data centers. With this central repository, further performance optimizations to VM storage data migration are made possible. Two migration mechanisms (data deduplication and Peer-to-Peer (P2P) file sharing) are utilized to accelerate base image migration, and a strategy is designed to elevate the synchronization of newly-written disk blocks. The results from an extensive experiment show that CBase significantly outperforms existing migration mechanisms under different conditions regarding total migration time and total network traffic. In particular, CBase with data deduplication is better than P2P file sharing for base image migration in our experimental environment. Highlights • A central repository is introduced to deploy and store base images for data centers. • Data deduplication and Peer-to-Peer (P2P) file sharing are used to accelerate the migration of a base image. • A mechanism is proposed to lower the overhead of migrating the writable layer of VM storage data. • An orchestration strategy is designed to improve the robustness of the migration system. • Experimental results validate the efficiency of our new approaches. [ABSTRACT FROM AUTHOR]

Abstract Edge computing is an emerging research area to incorporate cloud computing into edge network devices. An Edge datacenter, also referred to as EDC, processes data streams and user requests in real-time and is therefore used to decrease the latency and congestion in the network. EDC is usually setup as a distributed system and is accordingly placed between the cloud datacenter and the data source. These EDCs work as an intermediate layer in the fog hierarchy between IoT and Cloud datacenter. EDC's are aided by load balancers, responsible for distributing the workload amongst multiple EDC, in order to optimize resource utilization and response time. The load balancers make sure that the workload is equally divided amongst the available EDCs to avoid over loading of some EDCs while other remain idle as this directly impacts the user response and real-time event detection. Given the fact that EDCs are deployed in remote environments, the need for secure authentication is of major importance. In this paper we propose a novel load balancing technique that enables EDC authentication as well as identification of idle EDCs for better load balancing. The proposed load balancing technique is also compared with existing approaches and proves to be more efficient in locating EDC's with less workload. In addition to the improved efficiency, the proposed scheme also strengthens the security of the network by incorporating destination EDC authentication. Highlights • An adaptive EDC authentication technique with the help of a centralized cloud datacenter. • Authentication is initiated by the cloud and then all EDCs authenticate with each other. • Sustainable load balancing technique by considering the load of the destination EDCs. • The proposed approach combines both the authentication and load balancing technique. [ABSTRACT FROM AUTHOR]

Increased traffic demands within and between data centers now necessitate low-cost and low-power systems with per-wavelength bit rates beyond what can be easily achieved using conventional on/off keying. We review spectrally efficient links based on direct detection, Stokes vector detection, coherent detection and differentially coherent detection for data center applications. We show that limited spectral efficiency and power margin will inhibit scaling of direct detection-compatible formats beyond 100 Gbit/s. Stokes vector receivers can provide higher spectral efficiency without requiring a local oscillator laser, but require power-hungry analog-to-digital converters (ADCs) and digital signal processing (DSP). Similarly, existing DSP-based coherent systems designed for long-haul transmission may be excessively complex and power-hungry for short-reach data center links. We present low-power DSP-free coherent and differentially coherent alternatives that avoid high-speed ADCs and DSP and achieve similar performance to their DSP-based counterparts in intra-data center links and dispersion-compensated inter-data center links. [ABSTRACT FROM AUTHOR]

Increased demands for web and cloud-based services have been driving exponential growth of datacenter bandwidth. This paper discusses, from Google’s perspective, emerging challenges and possible technical solutions for scaling intra-datacenter and intra-campus interconnection network bandwidth. [ABSTRACT FROM AUTHOR]

On the basis of profound understanding of data center (DC) traffic demands and optical switching technologies, we present a hybrid optical network design for future data center network (DCN). Such design integrates optical circuit switching (OCS) and optical packet switching (OPS) schemes via hybrid Top-of-the-Rack (ToR) switches which provide flexible function-switchover between different traffic patterns in the DCN. Simulations of network behaviors under such DC traffic loads indicate that the proposed OPS/OCS DCN can effectively improve the network performance. Moreover, via a preliminary analysis of OCS and OPS network configurations, the construction of hybrid DCN is also proved as the most cost- and energy-efficient way for DCN upgrading while offering a promised quality of service. An experimental demonstration shows a complete implementation of data center virtualization in the proposed hybrid data center network. [ABSTRACT FROM AUTHOR]

Universal fiber is a multimode fiber that has an LP 01 mode field diameter approximately matched to that of standard single-mode fiber. It can transmit both multimode and single-mode signals using transceivers designed for either multimode fiber or single-mode fiber. By using universal fibers, one can bridge the needs for both single-mode and multimode transmission through a uniform and simplified cable infrastructure to accommodate the full distance range needed, while having upgradability from 10G to 40G to 100G and to even higher data rates. We build on our previous work that analyzed the dependence of LP 01 mode-field diameter on fiber parameters, and the coupling loss for both single-mode and multi-mode transmission, with a further study of mode coupling and multiple path interference associated with single-mode operation. We demonstrate system performance in both single-mode and VCSEL-based multimode transmissions for a number of 100G transceiver types using a QSFP form factor − the preferred choice in data centers. [ABSTRACT FROM AUTHOR]

This paper presents a novel VM allocation policy using the Eagle Strategy of Hybrid Krill Herd (KH) Optimization technique to enhance the cloud service experience to internet users. The Cloud Service provides data centre connectivity to share the resources through Virtual Machines (VM). The performance issue in Virtual Machine (VM) allocation during eccentrics degrades data centres load and clogging degrades the Quality of Experience (QoE) with Cloud Data Centres. Proposed optimized virtual machine allocation policy eradicates eccentrics effect and clogging effect, as it provides uninterrupted access throughput for better QoE. An optimization flag is additionally added in the Service Level Agreement (SLA). During burst load time, VM expansion is done automatically using Hybrid Krill Herd (KH) re-optimization if the optimization flag is set as true. Change and Response Protocol and Agreement Protocol have been proposed to predict the optimal resources for the Virtual Machine to avoid eccentrics and clogging. The prime parameters like delay, latency, packet rate and throughput are monitored continuously to activate the re-optimization and for achieving QoE with minimized load. Experimental analysis proves the capability of Hybrid KH algorithm over Particle Swarm Optimization, Ant Colony Optimization, Genetic Algorithm and Simulated Annealing algorithms. [ABSTRACT FROM AUTHOR]

Data center is the key infrastructure for a plenty of applications involving a high volume of data in Internet of Things (IoT). The Data Center Network (DCN) connecting multiple servers plays a vital role. Its mechanism for managing the traffic has a significant impact on the performance of IoT services. To guarantee the real-time performance of IoT services is one of the major challenges. In the paper Deadline-aware Rate Allocation (DRA) algorithm for scheduling the heterogeneous flows in DCNs is proposed. A non-cooperative game-theoretic framework is introduced to model the interactions in the scenario. The core idea of DRA is to assign the traffic with deadline constraints a higher priority. The worker with a lower served rate in the past period is assigned a higher priority. Meanwhile, DRA is a kind of preemptive algorithm. Simulation results have shown that under the mechanism flows wait shorter time and the real-time performance is guaranteed. DRA also achieves good fairness among different IoT services. [ABSTRACT FROM AUTHOR]

Energy management is becoming one of the major issues from last many years due to an exponential increase in the smart device users for accessing various services from the geo-distributed cloud data centers (DCs) using Internet. During this time, there is an emergence of new technology called as cloud computing which provides on-demand pay-per-use services such as storage, computation, and network to the end users. These services are provided to the end users by various geographically located DCs which are hosted by different service providers such as—Microsoft, Amazon, IBM etc. With an increase in dependence of end users, DCs have expanded both in size and number. With such an expansion, these DCs consume huge amount of energy to accomplish their routine tasks. Such an increase in energy consumption generates a lot of load on the power grid. Moreover, the carbon emissions from these DCs has a global impact on the environment. To mitigate these issues, the integration of DCs with renewable energy sources (RES) can be helpful to reduce the carbon emissions and to ease the load on the power grid. For this purpose, a SDN-based energy management scheme for sustainability of DCs using RES is proposed in this paper. 1 To achieve the aforementioned objectives, an energy-efficient flow scheduling algorithm is proposed using SDN. Moreover, a charging-discharging scheme for penetration of electric vehicles (EVs) is also presented to manage the intermittency of renewable energy. An energy trading and reward point scheme is designed to attract the EVs to participate in the proposed energy management scheme. The efficacy of the proposed scheme is proved using realistic weather traces. The results obtained clearly show the effectiveness of the proposed scheme for sustainability of DCs. [ABSTRACT FROM AUTHOR]