Geo-distributed Data Centers Research Articles

Towards Fault-Tolerant Federated and Distributed Machine Learning

Machine learning (ML) models are routinely trained and deployed among distributed devices, e.g., learning with geo-distributed data centers and federated learning with mobile devices. Such shared computing platforms are susceptible to hardware, software, communication errors, and security concerns. This talk will outline some of the threat models in distributed learning, along with robust learning methods proposed to augment the fault tolerance of distributed machine learning, showing both theoretical and empirical evidence of robustness to benign and adversarial attacks.

CloudSimPer: Simulating Geo-Distributed Datacenters Powered by Renewable Energy Mix

CloudSimPer: Simulating Geo-Distributed Datacenters Powered by Renewable Energy Mix

Joint optimization of operational cost and carbon emission in multiple data center micro-grids

As the internet data centers are mushrooming, the energy consumption and carbon emission of data centers are increasing rapidly. To cut down the electricity cost and carbon emission of the data centers, we proposed an optimization method to reduce the electricity cost and carbon emission in geo-distributed multiple data centers. In the proposed method, the carbon tax is considered in the overall operation cost to optimize the carbon emission. The spatial and temporal flexibility of computational workload is fully utilized by considering the difference in renewable energy power output, local electricity and carbon emission of multiple geo-distributed data centers to achieve a better performance. Furthermore, the nonlinear characteristics of the power loss of uninterruptible power supply (UPS) are considered in the optimization. To verify the proposed optimization method, simulation of six cases is carried out with realistic data, and results have proved the proposed method can reduce the operational costs by 4.93%–12.7% and decrease carbon emissions by up to 10%.

Uncertainty Level-Based Algorithms by Managing Renewable Energy for Geo-Distributed Datacenters

Uncertainty Level-Based Algorithms by Managing Renewable Energy for Geo-Distributed Datacenters

Towards Efficient Graph Processing in Geo-Distributed Data Centers

Towards Efficient Graph Processing in Geo-Distributed Data Centers

Geographically distributed data management to support large-scale data analysis

Nowadays, several companies prefer storing their data on multiple data centers with replication for many reasons. The data that spans various data centers ensures the fastest possible response time for customers and workforces who are geographically separated. It also provides protecting the information from the loss in case a single data center experiences a disaster. However, the amount of data is increasing at a rapid pace, which leads to challenges in storage, analysis, and various processing tasks. In this paper, we propose and design a geographically distributed data management framework to manage the massive data stored and distributed among geo-distributed data centers. The goal of the proposed framework is to enable efficient use of the distributed data blocks for various data analysis tasks. The architecture of the proposed framework is composed of a grid of geo-distributed data centers connected to a data controller (DCtrl). The DCtrl is responsible for organizing and managing the block replicas across the geo-distributed data centers. We use the BDMS system as the installed system on the distributed data centers. BDMS stores the big data file as a set of random sample data blocks, each being a random sample of the whole data file. Then, DCtrl distributes these data blocks into multiple data centers with replication. In analyzing a big data file distributed based on the proposed framework, we randomly select a sample of data blocks replicated from other data centers on any data center. We use simulation results to demonstrate the performance of the proposed framework in big data analysis across geo-distributed data centers.

Placement of High Availability Geo-Distributed Data Centers in Emerging Economies

The data center markets in emerging economies are being built at a furious pace. When high availability is required, as it always is in the modern digital economy, the placement of geo-distributed data centers may be influenced by factors such as technician shortage and under-developed infrastructure, both of which are typical in emerging economies. Although the data center availability subject in general has been well studied, it remains unclear how rapid and unbalanced economic development in emerging economies may affect the availability of geo-distributed data centers and their cost of ownership. In this paper, we incorporate the unbalanced availability of infrastructure and technician into the data center placement. The problem is first formulated as a mixed integer nonlinear program (MINLP).To solve this potentially large scale problem, we transform it into a QCQP, capable of handling heterogeneous workloads. The resulting problem can then be efficiently solved by off-the-shelf optimization toolboxes. With real-life data in China, we show how unbalanced development of infrastructure and technician shortage may affect the placement of data centers, and analyze the tradeoff between cost and availability. Our results indicate that technician shortage and unbalanced network infrastructure will lead to increased cost and distinct data center placement strategies.

Dual Optimization of Revenue and Expense in Geo-Distributed Data Centers Using Smart Grid

Exorbitant energy expenses can supersede data center profits. Electricity prices often vary across the geographic regions, caused by gaps in the supply-demand, time of use, and production cost factors. Geo-distributed cloud data centers facilitated by a smart grid and enabled by cloud computing can potentially utilize the spatiotemporal diversity of energy prices to reduce operational expenditure and maximize profit. In this paper, we solve the data center profit by formulating it as a constrained multi-objective optimization problem. The proposed solution utilizes an evolutionary algorithm-based higher-level heuristic that optimizes data center revenue and expense objectives simultaneously. The proposed technique provides system managers with trade-off solutions suited to varied operational scenarios. Ours is a multi-step approach, utilizing the optimization scheme to obtain Pareto optimal solutions for the request dispatch and resource allocation problem. When broadly evaluated against a comparative resource optimization scheme, our technique increases revenue while lowering expense and collectively yields a higher profit. It exhibits such performance over a broad range of price changes regardless of the data center's size and utilization level. The extensive simulation results ascertain the effectiveness of the proposed approach across a myriad of system parameters.

Energy Cost Minimization Using String Matching Algorithm in Geo-Distributed Data Centers

Data centers are being distributed worldwide by cloud service providers (CSPs) to save energy costs through efficient workload allocation strategies. Many CSPs are challenged by the significant rise in user demands due to their extensive energy consumption during workload processing. Numerous research studies have examined distinct operating cost mitigation techniques for geo-distributed data centers (DCs). However, operating cost savings during workload processing, which also considers string-matching techniques in geo-distributed DCs, remains unexplored. In this research, we propose a novel string matching-based geographical load balancing (SMGLB) technique to mitigate the operating cost of the geo-distributed DC. The primary goal of this study is to use a string-matching algorithm (i.e., Boyer Moore) to compare the contents of incoming workloads to those of documents that have already been processed in a data center. A successful match prevents the global load balancer from sending the user’s request to a data center for processing and displaying the results of the previously processed workload to the user to save energy. On the contrary, if no match can be discovered, the global load balancer will allocate the incoming workload to a specific DC for processing considering variable energy prices, the number of active servers, on-site green energy, and traces of incoming workload. The results of numerical evaluations show that the SMGLB can minimize the operating expenses of the geo-distributed data centers more than the existing workload distribution techniques.

Joint Workload Scheduling in Geo-Distributed Data Centers Considering UPS Power Losses

Recent years have witnessed the rapid growth in the electricity consumption of internet data centers (IDC) and consequently their high operational cost. Fortunately, geo-distributed multiple data centers could be jointly optimized to reduce the total operational cost using the flexible computational workload scheduling in terms of local electricity price and power consumption level of data center facilities, where in particular the power losses of uninterruptible power supply (UPS) contribute to around 5%. However, the nonlinear characteristics of UPS depending on load ratio have not been fully considered in the joint operation of multiple data centers. In view of this, this paper proposed a new joint workload scheduling method by carefully regulating the UPS power losses within multiple geo-distributed data centers. To precisely formulate the optimization problem, a comprehensive model of multiple geo-distributed data centers is built by considering local electricity price, renewable generation, conventional generator units, energy storage system (ESS), scheduling flexibility of different types of computational workload and especially the characteristics of UPSs. Furthermore, to guarantee the feasibility of the joint workload scheduling method when the renewable power and electricity varies, a day-ahead and intraday scheduling method is proposed for the first time. In consequence, six cases are simulated and solved by the Gurobi software, whose results have illustrated that the proposed method can reduce the operational cost by up to 4.69% under different scenarios.

Swing: Providing Long-Range Lossless RDMA via PFC-Relay

Remote Direct Memory Access (RDMA) has been widely deployed in datacenters for its high performance. Large-scale high performance cloud services built on geographically distributed datacenters require long-range RDMA for performance requirements. However, existing RDMA solutions can hardly satisfy the stringent requirements of the emerging large-scale high-performance cloud services built on geo-distributed datacenters in terms of throughput and delay. On the one hand, lossless RDMA suffers from a deep buffer and potential suboptimal throughput for inter-datacenter traffic due to delayed response to Priority Flow Control (PFC) messages. On the other hand, lossy RDMA with selective retransmissions suffers from poor performance when multiple flows with different round-trip times (RTTs) coexist in cross-datacenter scenarios. This article proposes <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Swing</small> , which expands the high-performance lossless RDMA to long-distance links through PFC-Relay. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Swing</small> ensures the throughput of long-distance links while minimizing the buffer requirement for long-range RDMA. It enables long-range RDMA without making any modifications to existing in-datacenter networks. The evaluation shows that <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Swing</small> can reduce the average flow completion time (FCT) by 14%-66% in a variety of traffic scenarios.

Workload and energy management of geo-distributed datacenters considering demand response programs

Datacenters usually consume huge energy and their workload can be adjusted dynamically, which makes them become good potential participants of demand response (DR) programs in the smart grid system. To overcome the drawbacks of centralization and enable fault-tolerance of a single datacenter, geographically distributed datacenters could be connected by a wide-area network to build up an integrated infrastructure, upon which multiple datacenter workloads could be balanced by taking advantage of space–time electricity price differences to reduce electricity costs. Particularly, when geo-distributed datacenters participate into DR programs, in order to obtain more economic benefits and to help achieve the load control objectives issued by the grid, it is necessary to consider a variety of external and internal conditions to optimize the workload and energy management strategy. Aiming at this problem, we propose an optimal power regulation scheduling method (OPRS) to simultaneously manage workload and energy consumption in geo-distributed datacenters, which comprehensively considers various factors including variation of electricity prices, operating costs, renewable energy constraints, and DR requirements. On the basis of the conventional computer room air conditioner system, additional devices are also incorporated into the architecture for datacenter cooling, including a solar-powered absorption cooling system and a chilled liquid storage system. Based on such framework, the OPRS method not only uses local renewable energy to reduce the energy consumption of the datacenter, but also can reduce the operating cost of the geo-distributed datacenters by leveraging the difference of electricity prices between time and space. Second, the method increases the frequency of uninterruptible power supply (UPS) usage and thus enhances the flexibility of datacenter power regulation. Furthermore, the OPRS method manages the workload by jointly considering task reassignment and delayed scheduling according to specific DR programs, thereby helping to maintain the stability of the regional power grid system. Experimental evaluation results show that, compared with other alternative strategies, the OPRS method can reduce the total operating cost by 19.52 %–66.96 %. Furthermore, the detailed analysis of the power regulation process illustrates that the OPRS method can also achieve better management of datacenter power consumption in the demand response stage.

DTC: A Dynamic Transaction Chopping Technique for Geo-Replicated Storage Services

Replicating data across geo-distributed datacenters is usually necessary for large scale cloud services to achieve high locality, durability and availability. One of the major challenges in such geo-replicated data services lies in consistency maintenance, which usually suffers from long latency due to costly coordination across datacenters. Among others, transaction chopping is an effective and efficient approach to address this challenge. However, existing chopping is conducted statically during programming, which is stubborn and complex for developers. In this article, we propose Dynamic Transaction Chopping (DTC), a novel technique that does transaction chopping and determines piecewise execution in a dynamic and automatic way. DTC mainly consists of two parts: a dynamic chopper to dynamically divide transactions into pieces according to the data partition scheme, and a conflict detection algorithm to check the safety of the dynamic chopping. Compared with existing techniques, DTC has several advantages: transparency to programmers, flexibility in conflict analysis, high degree of piecewise execution, and adaptability to data partition schemes. A prototype of DTC is implemented to verify the correctness of DTC and evaluate its performance. The experiment results show that our DTC technique can achieve much better performance than similar work.

MapReduce Task Scheduling in Heterogeneous Geo-Distributed Data Centers

Different data transmission times, processing times which are difficult to predict and node-dependent access times make MapReduce task scheduling rather complex. In this article, we consider the problem of scheduling MapReduce tasks to heterogeneous geo-distributed data centers to minimize the total tardiness. A new architecture is constructed to analyze data in the considered scenario. We model distinct data transmission levels, inter- and intra- data centers and heterogeneity of nodes mathematically. An algorithm framework is proposed to schedule MapReduce tasks to heterogeneous nodes in geographically distributed data centers. The proposed algorithm is suitable for both Hadoop MRv1 and MRv2. In terms of the number of idle containers detected in each heartbeat, the same number of tasks are selected from a sorted job sequence. For the map and reduce phases, two measurements are developed with data locality and completion time, respectively, based on which the classical Hungarian algorithm is adopted to optimally assign selected tasks to corresponding idle containers. Components and parameters of the proposal are statistically calibrated over a large set of random instances. A comparison of the proposed algorithm to existing methods for similar problems is carried out. Experimental results demonstrate the proposal is effective for the considered problem.

Renewable-Aware Geographical Load Balancing Using Option Pricing for Energy Cost Minimization in Data Centers

It is becoming increasingly difficult to properly control the power consumption of widely dispersed data centers. Energy consumption is high because of the need to run these data centers (DCs) that handle incoming user requests. The rising cost of electricity at the data center is a contemporary problem for cloud service providers (CSPs). Recent studies show that geo-distributed data centers may share the load and save money using variable power prices and pricing derivatives in the wholesale electricity market. In this study, we evaluate the problem of reducing energy expenditures in geographically dispersed data centers while accounting for variable system dynamics, power price fluctuations, and renewable energy sources. We present a renewable energy-based load balancing employing an option pricing (RLB-Option) online algorithm based on a greedy approach for interactive task allocation to reduce energy costs. The basic idea of RLB-Option is to process incoming user requests using available renewable energy sources. In contrast, in the case of unprocessed user requests, the workload will be processed using brown energy or call option contract at each timeslot. We formulate the energy cost minimization in geo-distributed DCs as an optimization problem considering geographical load balancing, renewable energy, and an option pricing contract from the derivative market while satisfying the set of constraints. We prove that the RLB-Option can reduce the energy cost of the DCs close to that of the optimal offline algorithm with future information. Compared to standard workload allocation methods, RLB-Option shows considerable cost savings in experimental evaluations based on real-world data.

Truthful auction mechanisms for VNF chain provisioning and allocation across geo-distributed datacenters

With the fast development of network function virtualization (NFV), NFV markets are emerging that allows network service providers (NSPs) to trade different virtual network function (VNF) chains among users. Therefore, efficient mechanisms for VNF chain provisioning and allocation are key to guarantee efficient operations of the NFV markets. One fundamental issue is how to maximize the social welfare of the market such that selfish users have incentives to acquire VNF chains from the markets. To overcome this issue, in this paper, we propose a truthful randomized combinatorial auction mechanism (TRCAM), which integrates the primal–dual approximation algorithm design technique and efficient pricing strategy. This mechanism considers dynamic VNF chain provisioning of the NSP and flexible VNF chain deployment of users. To be specific, we first formulate the social welfare maximization problem as an integer linear program, which is proven to be NP-hard. Then, we propose a primal–dual approximate algorithm which is near-optimal with polynomial-time complexity and employ it as a building block to design the TRCAM. Furthermore, we develop a novel binary-search-based algorithm to boost the average-case performance of the TRCAM. Both strict theoretical analysis and extensive experimental studies demonstrate the effectiveness of our proposed TRCAM, which achieves approximate truthfulness, near-optimal social welfare, individual rationality, and computational efficiency.

Profit Maximization in a Federated Cloud by Optimal Workload Management and Server Speed Setting

In an environment with multiple heterogeneous multiservers across multiple clouds and data centers, optimal workload management and server speed setting makes strong impact on the aggregated performance, cost, and profit of a cloud service provider. Such a situation provides us a more challenging opportunity to address and discuss profit maximization of a service provider in a wider scale and to generate more significant influence. In this paper, we investigate profit maximization by optimal workload management and server speed setting for multiple heterogeneous multiservers in a federated cloud or a geo-distributed data center. The heterogeneous multiservers have different sizes, speeds, power consumption models, workload, performance, costs, and profit. To conduct rigorous study, each multiserver system is modeled by an M/M/m queueing system, such that the profit of a multiserver system can be characterized analytically. We address two problems, i.e., workload management without server speed setting and workload management with server speed setting. Both problems are formulated as multi-variable optimization problems. We develop numerical algorithms to solve these problems. We also provide numerical data for the purpose of illustration. To the best of the author's knowledge, this is the first paper which analytically discusses profit maximization for multiple heterogeneous multiservers in a federated cloud or a geo-distributed data center using queueing systems.

Big Data Processing in Geo-Distributed Data Centers Using Cost Minimization Methods

Abstract: The explosive growth of demands on huge process imposes a crucial burden on computation, storage, and communication in data centers, that so incurs wide operational expenditure to data center suppliers. Therefore, Price decrease has become Associate in Nursing rising issue for the longer term huge data era. All totally different from customary cloud services, one in each of the foremost choices of huge data services is that the tight coupling between data and computation as computation tasks is also conducted provided that the corresponding data is obtainable. As a result, 3 factors, i.e., task assignment, data placement and data movement, deeply the operational expenditure of data centers. Throughout this paper, we tend to tend to unit of measurement meant to see the worth decrease downside via a joint improvement of these three factors for big data services in geo-distributed data centers. To elucidate the task completion time with the thought of every data transmission and computation, we tend to tend to propose a two-dimensional stochastic process and derive the standard task completion time in closed-form. Moreover, we tend to tend to model the matter as a mixed-integer non-linear programming (MINLP) and propose economical resolution to correct it. The high efficiency of our proposal is valid by thorough simulation primarily based studies Keywords— Nonlinear Programming, Big Data, Computer Centres, Integer Programming, Minimisation, Mixed Integer Nonlinear Programming, Cost Minimization, Data Centers, Big Data Services, Operational Expenditure, Task Assignment, Big Data

EDCSuS: Sustainable Edge Data Centers as a Service in SDN-Enabled Vehicular Environment

Cloud computing has emerged as one of the popular technologies which provide on-demand services to the end users. Such services are hosted by massive geo-distributed data centers (DCs). Nowadays, connected vehicles in a smart city can also avail cloud services through Internet using cellular technologies. But, the advent of 5G technology has posed challenges for DCs such as-low latency and higher data rate requirements. To handle these challenges, edge-DCs (EDCs) can be deployed across a smart city to provide low latency services to the connected vehicles. In lieu of this, in this paper, EDCSuS: Sustainable EDC as a service framework in software defined vehicular environment is proposed. In EDCSuS, first, a software defined controller handles the incoming requests and suggest an optimal flow path. Second, a multi-leader multi-follower Stackelberg game is presented for resource allocation. Third, to improve the resource utilization, a cooperative resource sharing scheme is designed, thereby minimizing the energy consumption of servers in the EDCs. Lastly, a caching scheme is presented to avert excessive energy consumption for retracing the lost link due to vehicular mobility. The efficacy of the proposed scheme has been evaluated using extensive simulations with respect to various parameters. The results obtained prove the effectiveness of EDCSuS.

Energy and Network Aware Workload Management for Geographically Distributed Data Centers

Cloud service providers are distributing data centers geographically to minimize energy costs through intelligent workload distribution. With increasing data volumes in emerging cloud workloads, it is critical to factor in the network costs for transferring workloads across data centers. For geo-distributed data centers, many researchers have been exploring strategies for energy cost minimization and intelligent inter-data-center workload distribution separately. However, prior work does not comprehensively and simultaneously consider data center energy costs, data transfer costs, and data center queueing delay. In this paper, we propose a novel game theory-based workload management framework that takes a holistic approach to the cloud operating cost minimization problem by making intelligent scheduling decisions aware of data transfer costs and the data center queueing delay. Our framework performs intelligent workload management that considers heterogeneity in data center compute capability, cooling power, interference effects from task co-location in servers, time-of-use electricity pricing, renewable energy, net metering, peak demand pricing distribution, and network pricing. Our simulations show that the proposed game-theoretic technique can minimize the cloud operating cost more effectively than existing approaches.