Search

Big scale, high performance and fault-tolerance, low-cost and graceful expandability are pursued features in current datacenter networks (DCN). Although there have been many proposals for DCNs, most modern installations are equipped with classical folded Clos networks. Recently, regular random topologies, as the Jellyfish, have been proposed for DCNs. However, their completely unstructured nature entails serious design problems. In this paper we propose Random Folded Clos (RFC) and Hydra networks in which the interconnection between certain switches levels is made randomly. Both RFCs and Hydras preserve important properties of Clos networks that provide a straightforward deadlock-free multi-path routing. The proposed networks leverage randomness to be gracefully expandable, thereby allowing for fine grain upgrading. RFCs and Hydras are compared in the paper, in topological and cost terms, against fat-trees, orthogonal fat-trees and random regular networks. Also, experiments are carried out to simulate their performance under synthetic traffic patterns emulating common loads present in warehouse scale computers. These theoretical and empirical studies reveal the interest of these topologies, concluding that Hydra constitutes a practicable alternative to current datacenter networks since it appropriately balance all the main design requirements. Moreover, Hydras perform better than the fat-trees, their natural competitor, being able to connect the same or more computing nodes with significant lower cost and latency while exhibiting comparable throughput. [ABSTRACT FROM AUTHOR]

As the volume of data keeps growing rapidly, more and more storage devices, servers and network devices are continuously added into data centers to store, manage and analyze the data. The industry experience indicates that, instead of a huge number of servers added at a time, the data center network also expands gradually by adding a small number of servers from time to time. As a result, how to achieve an incremental scalability is becoming a very important challenge in designing modern data center network architectures in order to maintain the topological properties unchanged when the size of data centers grows. In this paper, we propose a new type of data center network architecture named HSDC (High Scalability Data Center Network Architecture) based on the hypercube network. The HSDC is constructed by using $m$m-port switches and 2-port servers. The fault-tolerant routing algorithm designed in this paper for HSDC can be executed on any vertex and is able to construct a path between any pair of vertices. In order to achieve an incremental scalability, we further propose three types of incomplete HSDC structures that allow gradually adding servers into the structures, while maintaining all the topological properties. The simulation experiments and performance results demonstrate that the throughput of HSDC is comparable to that of Fat-Tree, BCube and DCell. Furthermore, the analysis results indicate that HSDC strikes a good balance among diameter, bisection width, incremental scalability, cost and energy consumption in contrast to the state-of-the-art data center network architectures. [ABSTRACT FROM AUTHOR]

Today’s data center operators deploy network policies in both physical (e.g., middleboxes, switches) and virtualized (e.g., virtual machines on general purpose servers) network function boxes (NFBs), which reside in different points of the network, to exploit their efficiency and agility respectively. Nevertheless, such heterogeneity has resulted in a great number of independent network nodes that can dynamically generate and implement inconsistent and conflicting network policies, making correct policy implementation a difficult problem to solve. Since these nodes have varying capabilities, services running atop are also faced with profound performance unpredictability. In this paper, we propose a Heterogeneous netwOrk Policy Enforcement (HOPE) scheme to overcome these challenges. HOPE guarantees that network functions (NFs) that implement a policy chain are optimally placed onto heterogeneous NFBs such that the network cost of the policy is minimized. We first experimentally demonstrate that the processing capacity of NFBs is the dominant performance factor. This observation is then used to formulate the Heterogeneous Network Policy Placement problem, which is shown to be NP-Hard. To solve the problem efficiently, an online algorithm is proposed. Our experimental results demonstrate that HOPE achieves the same optimality as Branch-and-bound optimization but is 3 orders of magnitude more efficient. [ABSTRACT FROM AUTHOR]

GPUs have been becoming an indispensable computing platform in data centers, and co-locating multiple applications on the same GPU is widely used to improve resource utilization. However, performance interference due to uncontrolled resource contention severely degrades the performance of co-locating applications and fails to deliver satisfactory user experience. In this paper, we present SMGuard, a software approach to flexibly manage the GPU resource usage of multiple applications under co-location. We also propose a capacity based GPU resource model CapSM, which provisions the GPU resources in a fine-grained granularity among co-locating applications. When co-locating latency-sensitive applications with batch applications, SMGuard can prevent batch applications from occupying resources without constraint using quota based mechanism, and guarantee the resource usage of latency-sensitive applications with reservation based mechanism. In addition, SMGuard supports dynamic resource adjustment through evicting the running thread blocks of batch applications to release the occupied resources and remapping the uncompleted thread blocks to the remaining resources, which avoids the relaunch of the preempted kernel. The SMGuard is a pure software solution that does not rely on special GPU hardware or programming model, which is easy to adopt on commodity GPUs in data centers. Our evaluation shows that SMGuard improves the average performance of latency-sensitive applications by 9.8× when co-located with batch applications. In the meanwhile, the GPU utilization can be improved by 35 percent on average. [ABSTRACT FROM AUTHOR]

Power optimization has become a key challenge in the design of large-scale enterprise data centers. Existing research efforts focus mainly on computer servers to lower their energy consumption, while only few studies have tried to address data center networks (DCNs), which can account for 10-20 percent of the total energy consumption of a data center. In this paper, we propose CARPO, a correlation-aware power optimization algorithm that dynamically consolidates traffic flows onto a small set of links and switches in a DCN and then shuts down unused network devices for energy savings. In sharp contrast to existing work, CARPO is designed based on a key observation from the analysis of real DCN traces that the bandwidth demands of different flows do not peak at exactly the same time. As a result, if the correlations among flows are considered in consolidation, more energy savings can be achieved. In addition, CARPO integrates traffic consolidation with link rate adaptation for maximized energy savings. We implement CARPO on a hardware testbed composed of 10 virtual switches configured with a production 48-port OpenFlow switch and 8 servers. Our empirical results with traces from Wikipedia and Yahoo! data centers demonstrate that CARPO can save up to 50 percent of network energy for a DCN, while having only negligible delay increases. CARPO also outperforms two state-of-the-art baselines by 19.6 and 95 percent on energy savings, respectively. Our simulation results with a large-scale DCN also show that CARPO can achieve more energy savings than the baselines for typical DCN topologies, such as fat tree and BCube. [ABSTRACT FROM PUBLISHER]

The end of Dennard's scaling poses computer systems, especially the datacenters, in front of both power and utilization walls. One possible solution to combat the power and utilization walls is dark silicon where transistors are under-utilized in the chip, but this will result in a diminishing performance. Another solution is Near-Threshold Voltage Computing (NTC) which operates transistors in the near-threshold region and provides much more flexible tradeoffs between power and performance. However, prior efforts largely focus on a specific design option based on the legacy desktop applications, therefore, lacking comprehensive analysis of emerging scale-out applications with multiple design options when dark silicon and/or NTC are/is applied. In this paper, we characterize different perspectives including performance, energy efficiency and reliability in the context of NTC/dark silicon cloud processors running emerging scale-out workloads on various architecture designs. We find NTC is generally an effective way to alleviate the power challenge over scale-out applications compared with dark silicon, it can improve performance by 1.6X, energy efficiency by 50 percent and the reliability problem can be relieved by ECC. Meanwhile, we also observe tiled-OoO architecture improves the performance by 20∼370 percent and energy efficiency by 40 ∼ 600 percent over alternative architecture designs, making it a preferable design paradigm for scale-out workloads. We believe that our observations will provide insights for the design of cloud processors under dark silicon and/or NTC . [ABSTRACT FROM AUTHOR]

Modern day data centers coordinate hundreds of thousands of heterogeneous tasks and aim at delivering highly reliable cloud computing services. Although offering equal reliability to all users benefits everyone at the same time, users may find such an approach either inadequate or too expensive to fit their individual requirements, which may vary dramatically. In this paper, we propose a novel method for providing elastic reliability optimization in cloud computing. Our scheme makes use of peer-to-peer checkpointing and allows user reliability levels to be jointly optimized based on an assessment of their individual requirements and total available resources in the data center. We show that the joint optimization can be efficiently solved by a distributed algorithm using dual decomposition. The solution improves resource utilization and presents an additional source of revenue to data center operators. Our validation results suggest a significant improvement of reliability over existing schemes. [ABSTRACT FROM PUBLISHER]

Virtualization of servers and networks is a key technique to resolve the conflict between the increasing demands on computing power and the high cost of hardware in data centers. In order to map virtual networks to physical infrastructure efficiently, designers have to make careful decisions on the allocation of limited resources, which makes placement of virtual networks in data centers a critical issue. In this paper, we study the placement of virtual networks in fat-tree data center networks. In order to meet the requirements of instant parallel data transfer between multiple computing units, we propose a model of multicast-capable virtual networks (MVNs). We then design four virtual machine (VM) placement schemes to embed MVNs into fat-tree data center networks, named Most-Vacant-Fit (MVF), Most-Compact-First (MCF), Mixed-Bidirectional-Fill (MBF), and Malleable-Shallow-Fill (MSF). All these VM placement schemes guarantee the nonblocking multicast capability of each MVN while simultaneously achieving significant saving in the cost of network hardware. In addition, each VM placement scheme has its unique features. The MVF scheme has zero interference to existing computing tasks in data centers; the MCF scheme leads to the greatest cost saving; the MBF scheme simultaneously possesses the merits of MVF and MCF, and it provides an adjustable parameter allowing cloud providers to achieve preferred balance between the cost and the overhead; the MSF scheme performs at least as well as MBF, and possesses some additional predictable features. Finally, we compare the performance and overhead of these VM placement schemes, and present simulation results to validate the theoretical results. [ABSTRACT FROM AUTHOR]

In this paper, we consider the design space of parallel non-blocking slot scheduling algorithms. Slot schedulers divide time into discrete quanta called slots, and schedule resources at the granularity of slots. They are typically used in high throughput I/O systems, data centers, video servers, and network drivers. We propose a family of parallel slot scheduling problems of increasing complexity, and then propose parallel lock-free and wait-free algorithms to solve them. In specific, we propose problems that can reserve, as well as free a set of contiguous slots in a non-blocking manner. We show that in a system with 64 threads, it is possible to get speedups of 10X by using lock-free algorithms as compared to a baseline implementation that uses locks. We additionally propose wait-free algorithms, whose mean performance is roughly the same as the version with locks. However, they suffer from significantly lower jitter and ensure a high degree of fairness among threads. [ABSTRACT FROM PUBLISHER]

Cloud computing applications face the challenges of dealing with a huge volume of data that needs the support of accurate and fast approximate queries to enhance system scalability and improve quality of service. Locality-sensitive hashing (LSH) can support the approximate queries that unfortunately suffer from imbalanced load and space inefficiency among distributed data servers, which severely limits the query accuracy and incurs long query latency between users and cloud servers. In this paper, we propose a novel scheme, called NEST, which offers easy-to-use and cost-effective approximate queries for cloud computing. The novelty of NEST is to leverage cuckoo-driven locality-sensitive hashing to find similar items that are further placed closely through cuckoo-driven method to obtain load-balancing buckets in hash tables. NEST hence carries out flat and manageable addressing in adjacent buckets, and obtains constant-scale query complexity even in the worst case. The benefits of NEST include the increments of space utilization and fast query response. Moreover, due to the salient property of flat addressing in NEST, we implement NEST design in a real hybrid storage system, which consists of DRAM, SSD, and hard disk. The flat addressing allows efficient operations in SSD to improve system performance. We argue that a proper “ division of labor” among DRAM, SSD, and hard disk in the hybrid and heterogeneous storage hierarchy is desperately needed to strike an optimal balance to remove the indexing bottleneck. Theoretical analysis and extensive experiments (on LANL and Microsoft metadata) in a large-scale cloud testbed demonstrate the salient properties of NEST to meet the needs of approximate query service in cloud computing environments. We have offered open-source codes of NEST for public use. [ABSTRACT FROM PUBLISHER]

Traffic in operator networks is time varying. Conventional network functions implemented by black-boxes should satisfy the peak traffic requirement, and hence result in low resource utilization. Thanks to the emergence of Virtual Network Function (VNF), which is realized by running networking software on Virtual Machines (VMs), the operator can dynamically scale in or scale out the VNF instances and hence save the required resources. In this paper, we introduce how the dynamic VNF scaling is implemented in practical operator Data Center Networks (DCNs). First, we analyze the traffic characteristics in our operator networks, and introduce how the VNFs are organized in a common operator DCN. Based on these backgrounds, we not only propose a traffic forecasting method, but also design two VNF placement algorithms to guide the dynamic VNF instance scaling. Through both the implementation in a real operator network and extensive real trace driven simulations, we demonstrate that our dynamic VNF instance scaling system can achieve higher service availability and save the VNF resources (e.g., CPU and memory) by up to 30 percent. [ABSTRACT FROM AUTHOR]

The fast evolution of data analytics platforms has resulted in an increasing demand for real-time data stream processing. From Internet of Things applications to the monitoring of telemetry generated in large data centers, a common demand for currently emerging scenarios is the need to process vast amounts of data with low latencies, generally performing the analysis process as close to the data source as possible. Stream processing platforms are required to be malleable and absorb spikes generated by fluctuations of data generation rates. Data is usually produced as time series that have to be aggregated using multiple operators, being sliding windows one of the most common abstractions used to process data in real-time. To satisfy the above-mentioned demands, efficient stream processing techniques that aggregate data with minimal computational cost need to be developed. In this paper we present the Monoid Tree Aggregator general sliding window aggregation framework, which seamlessly combines the following features: amortized $O(1)$ time complexity and a worst-case of $O(\log {n})$ between insertions; it provides both a window aggregation mechanism and a window slide policy that are user programmable; the enforcement of the window sliding policy exhibits amortized $O(1)$ computational cost for single evictions and supports bulk evictions with cost $O(\log {n})$ ; and it requires a local memory space of $O(\log {n})$ . The framework can compute aggregations over multiple data dimensions, and has been designed to support decoupling computation and data storage through the use of distributed Key-Value Stores to keep window elements and partial aggregations. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the problem of minimizing the long-term energy cost for an Internet data center (IDC) in deregulated electricity markets. Specifically, IDC operators intend to minimize energy cost by scheduling workload and battery jointly, which can fully exploit the temporal diversity of electricity price. First, we formulate a stochastic optimization problem takingheterogeneous service delay guarantees and battery management into account. Then, we design an online operation algorithm to solve the problem based on Lyapunov optimization technique. Meanwhile, we analyze the feasibility and performance guarantee of the proposed algorithm. Finally, extensive simulation results based on real-world data show the effectiveness of the proposed algorithm. [ABSTRACT FROM PUBLISHER]

Lately, for geo-distributed data centers, a workload management approach that routes user requests to locations with cheaper and cleaner electricity has been developed to reduce energy consumption and cost. We consider two key aspects that have not been explored in this approach. First, through empirical studies, we find that the energy efficiency of cooling systems depends critically on the ambient temperature, which exhibits significant geographical diversity. Temperature diversity can be used to reduce the cooling energy overhead. Second, energy consumption comes from not only interactive workloads driven by user requests, but also delay tolerant batch workloads that run at the back-end. The elastic nature of batch workloads can be exploited to further reduce the energy cost. In this paper, we propose to make workload management temperature aware. We formulate the problem as a joint optimization of request routing for interactive workloads and capacity allocation for batch workloads. We develop a distributed algorithm based on an $m$ -block alternating direction method of multipliers (ADMM) algorithm that extends the classical two-block algorithm. We prove the convergence and rate of convergence results under general assumptions. Through trace-driven simulations, we find that our approach consistently provides 15-20 percent cooling energy reduction, and 5-20 percent overall cost reduction over existing methods. [ABSTRACT FROM PUBLISHER]

Fat trees are considered suitable structures for data center interconnection networking. Such structures are rigid, and hard to scale up and scale out. A good data center network structure should have high scalability, efficient switch utilization, and high reliability. In this paper we present a class of data center network structures based on hypergraph theory and combinatorial block design theory. We show that our data center network structures are more flexible and scalable than fat trees. Using switches of the same size, our data center network structures can connect more nodes than fat trees, and it is possible to construct different structures with tradeoffs among inter-cluster communication capacity, reliability, the number of switches used, and the number of connected nodes. [ABSTRACT FROM AUTHOR]

Distributed computing systems like MapReduce in data centers transfer massive amount of data across successive processing stages. Such shuffle transfers contribute most of the network traffic and make the network bandwidth become a bottleneck. In many commonly used workloads, data flows in such a transfer are highly correlated and aggregated at the receiver side. To lower down the network traffic and efficiently use the available network bandwidth, we propose to push the aggregation computation into the network and parallelize the shuffle and reduce phases. In this paper, we first examine the gain and feasibility of the in-network aggregation with BCube, a novel server-centric networking structure for future data centers. To exploit such a gain, we model the in-network aggregation problem that is NP-hard in BCube. We propose two approximate methods for building the efficient IRS-based incast aggregation tree and SRS-based shuffle aggregation subgraph, solely based on the labels of their members and the data center topology. We further design scalable forwarding schemes based on Bloom filters to implement in-network aggregation over massive concurrent shuffle transfers. Based on a prototype and large-scale simulations, we demonstrate that our approaches can significantly decrease the amount of network traffic and save the data center resources. Our approaches for BCube can be adapted to other server-centric network structures for future data centers after minimal modifications. [ABSTRACT FROM AUTHOR]

Along with the rapid increasing energy consumption, the energy cost of Internet data centers (IDCs) has been skyrocketing. A novel scheme of geographical load balancing was proposed to reduce electricity bills for service providers. However, one important challenge faced by service providers has not been considered properly. In service systems, the service delay faced by consumers includes the queuing delay and the transmission delay. While existing work only consider the queuing delay, the transmission delay introduced by geographical load balancing has been overlooked. It is one of the most important factors affecting the quality of service for real-time service systems. In this paper, we take the transmission delay into our design consideration and formulate a mixed-integer nonlinear programming (MINLP) problem with coupled constraint to achieve the optimal load balancing and energy cost management for IDCs while meeting the service-level agreements (SLA) of consumers. A novel heuristic based branch and bound with feedback (HBBF) algorithm is proposed to decouple the MINLP problem with coupled constraint efficiently. Extensive performance evaluations based on real electricity price data and site-to-site transmission delay data demonstrate the effectiveness of our proposed algorithm. [ABSTRACT FROM AUTHOR]

Cloud computing services are becoming integral part of people's daily life. These services are supported by infrastructure known as Internet data center (IDC). As demand for cloud computing services soars, energy consumed by IDCs is skyrocketing. Both academia and industry have paid great attention to energy management of IDCs. This paper studies an important energy management problemâhow to minimize energy cost for IDCs in deregulated electricity markets. We propose a novel two-stage design and the eco-IDC (Energy Cost Optimization-IDC) algorithm to exploit the temporal diversity of electricity price and dynamically schedule workload to execute on IDC servers through an input queue. Extensive evaluation experiments are performed using real-life electricity price and workload traces at an enterprise production data center. The evaluation results demonstrate that the proposed approach significantly reduces energy cost for IDCs, guarantees a service delay bound, and alleviates workload drop if the service delay bound is sufficiently large. [ABSTRACT FROM AUTHOR]

Network function virtualization (NFV) brings great conveniences and benefits for the enterprises to outsource their network functions to the cloud datacenter. In this paper, we address the virtual network function (VNF) placement problem in cloud datacenter considering users’ service function chain requests (SFCRs). To optimize the resource utilization, we take two less-considered factors into consideration, which are the time-varying workloads, and the basic resource consumptions (BRCs) when instantiating VNFs in physical machines (PMs). Then the VNF placement problem is formulated as an integer linear programming (ILP) model with the aim of minimizing the number of used PMs. Afterwards, a Two-StAge heurisTic solution (T-SAT) is designed to solve the ILP. T-SAT consists of a correlation-based greedy algorithm for SFCR mapping (first stage) and a further adjustment algorithm for virtual network function requests (VNFRs) in each SFCR (second stage). Finally, we evaluate T-SAT with the artificial data we compose with Gaussian function and trace data derived from Google's datacenters. The simulation results demonstrate that the number of used PMs derived by T-SAT is near to the optimal results and much smaller than the benchmarks. Besides, it improves the network resource utilization significantly. [ABSTRACT FROM AUTHOR]

Lambda paradigm, also known as Function as a Service (FaaS), is a novel event-driven concept that allows companies to build scalable and reliable enterprise applications in an off-premise computing data-center as a serverless solution. In practice, however, an important goal for the service provider of a Lambda platform is to devise an efficient way to consolidate multiple Lambda functions in a single host. While the majority of existing resource management solutions use only operating-system level metrics (e.g., average utilization of computing and I/O resources) to allocate the available resources among the submitted workloads in a balanced way, a resource allocation schema that is oblivious to the issue of shared-resource contention can result in a significant performance variability and degradation within the entire platform. This paper proposes a predictive controller scheme that dynamically allocates resources in a Lambda platform. This scheme uses a prediction tool to estimate the future rate of every event stream and takes into account the quality of service enforcements requested by the owner of each Lambda function. This is formulated as an optimization problem where a set of cost functions are introduced (i) to reduce the total QoS violation incidents; (ii) to keep the CPU utilization level within an accepted range; and (iii) to avoid the fierce contention among collocated applications for obtaining shared resources. Performance evaluation is carried out by comparing the proposed solution with an enhanced interference-aware version of three well-known heuristics, namely spread, binpack (the two native clustering solutions employed by Docker Swarm) and best-effort resource allocation schema. Experimental results show that the proposed controller improves the overall performance (in terms of reducing the end-to-end response time) by 14.9 percent on average compared to the best result of the other heuristics. The proposed solution also increases the overall CPU utilization by 18 percent on average (for lightweight workloads), while achieves an average 87 percent (maximum 146 percent) improvement in preventing QoS violation incidents. [ABSTRACT FROM AUTHOR]

Increasing number of Internet-scale applications, such as video streaming, incur huge amount of wide area traffic. Such traffic over the unreliable Internet without bandwidth guarantee suffers unpredictable network performance. This result, however, is unappealing to the application providers. Fortunately, Internet giants like Google and Microsoft are increasingly deploying their private wide area networks (WANs) to connect their global datacenters. Such high-speed private WANs are reliable, and can provide predictable network performance. In this paper, we propose a new type of service—inter-datacenter network as a service (iDaaS), where traditional application providers can reserve bandwidth from those Internet giants to guarantee their wide area traffic. Specifically, we design a bandwidth trading market among multiple iDaaS providers and application providers, and concentrate on the essential bandwidth pricing problem. The involved challenging issue is that the bandwidth price of each iDaaS provider is not only influenced by other iDaaS providers, but also affected by the application providers. To address this issue, we characterize the interaction between iDaaS providers and application providers using a Stackelberg game model, and analyze the existence and uniqueness of the equilibrium. We further present an efficient bandwidth pricing algorithm by blending the advantage of a geometrical Nash bargaining solution and the demand segmentation method. For comparison, we present two bandwidth reservation algorithms, where each iDaaS provider's bandwidth is reserved in a weighted fair manner and a max-min fair manner, respectively. Finally, we conduct comprehensive trace-driven experiments. The evaluation results show that our proposed algorithms not only ensure the revenue of iDaaS providers, but also provide bandwidth guarantee for application providers with lower bandwidth price per unit. [ABSTRACT FROM AUTHOR]

In the cloud, multiple tenants share the resource of datacenters and their applications compete with each other for scarce network bandwidth. Current studies have shown that the lack of bandwidth guarantee causes unpredictable network performance, leading to poor application performance. To address this issue, several virtual network abstractions have been proposed which allow the tenants to reserve virtual clusters with specified bandwidth between the Virtual Machines (VMs) in the datacenters. However, all these existing proposals require the tenants to deterministically characterize the bandwidth demands in the abstractions, which can be difficult and result in inefficient bandwidth reservation due to the demand uncertainty. In this paper, we explore a virtual cluster abstraction with stochastic bandwidth characterization to address the bandwidth demand uncertainty. We propose Stochastic Virtual Cluster (SVC), which models the bandwidth demand between VMs in a probabilistic way. Based on SVC, we develop a stochastic framework for virtual cluster allocation, in which the admitted virtual cluster's bandwidth demands are satisfied with a high probability. Efficient VM allocation algorithms are proposed to implement the framework while reducing the possibility of link congestion through minimizing the maximum bandwidth occupancy of a virtual cluster on physical links. Using simulations, we show that SVC achieves the trade-off between the job concurrency and the average job running time, and demonstrate its effectiveness for accommodating cloud application workloads with highly volatile bandwidth demands and its improvement to work-conserving bandwidth enforcement. [ABSTRACT FROM AUTHOR]

Many power management strategies have been proposed for enterprise servers based on dynamic voltage and frequency scaling (DVFS), but those solutions cannot further reduce the energy consumption of a server when the server processor is already at the lowest DVFS level and the server utilization is still low (e.g., 10 percent or lower). To achieve improved energy efficiency, request batching can be conducted to group received requests into batches and put the processor into sleep between the batches. However, it is challenging to perform request batching on a virtualized server because different virtual machines on the same server may have different workload intensities. Hence, putting the shared processor into sleep may severely impact the application performance of all the virtual machines. This paper proposes Virtual Batching, a novel request batching solution for virtualized servers with primarily light workloads. Our solution dynamically allocates CPU resources such that all the virtual machines can have approximately the same performance level relative to their allowed peak values. Based on this uniform level, Virtual Batching determines the time length for periodically batching incoming requests and putting the processor into sleep. When the workload intensity changes from light to moderate, request batching is automatically switched to DVFS to increase processor frequency for performance guarantees. Virtual Batching is also extended to integrate with server consolidation for maximized energy conservation with performance guarantees for virtualized data centers. Empirical results based on a hardware testbed and real trace files show that Virtual Batching can achieve the desired performance with more energy conservation than several well-designed baselines, e.g., 63 percent more, on average, than a solution based on DVFS only. [ABSTRACT FROM PUBLISHER]

Energy consumption represents a significant cost in data center operation. A large fraction of the energy, however, is used to power idle servers when the workload is low. Dynamic provisioning techniques aim at saving this portion of the energy, by turning off unnecessary servers. In this paper, we explore how much gain knowing future workload information can bring to dynamic provisioning. In particular, we develop online dynamic provisioning solutions with and without future workload information available. We first reveal an elegant structure of the offline dynamic provisioning problem, which allows us to characterize the optimal solution in a “divide-and-conquer” manner. We then exploit this insight to design two online algorithms with competitive ratios $(2-\alpha)$ and $(e/(e-1+\alpha ))$, respectively, where $(0\le \alpha \le 1)$ is the normalized size of a look-ahead window in which future workload information is available. A fundamental observation is that future workload information beyond the full-size look-ahead window (corresponding to $(\alpha =1)$) will not improve dynamic provisioning performance. Our algorithms are decentralized and easy to implement. We demonstrate their effectiveness in simulations using real-world traces. [ABSTRACT FROM AUTHOR]

Electricity expenditure comprises a significant fraction of the total operating cost in data centers. Hence, cloud service providers are required to reduce electricity cost as much as possible. In this paper, we consider utilizing existing energy storage capabilities in data centers to reduce electricity cost under wholesale electricity markets, where the electricity price exhibits both temporal and spatial variations. A stochastic program is formulated by integrating the center-level load balancing, the server-level configuration, and the battery management while at the same time guaranteeing the quality-of-service experience by end users. We use the Lyapunov optimization technique to design an online algorithm that achieves an explicit tradeoff between cost saving and energy storage capacity. We demonstrate the effectiveness of our proposed algorithm through extensive numerical evaluations based on real-world workload and electricity price data sets. As far as we know, our work is the first to explore the problem of electricity cost saving using energy storage in multiple data centers by considering both the spatial and temporal variations in wholesale electricity prices and workload arrival processes. [ABSTRACT FROM AUTHOR]

Cloud data centers are widely employed to offer reliable cloud services. However, low resource utilization and high power consumption have been great challenges for cloud providers. Moreover, the rapid increase in demand for affordable cloud services magnifies the obstacles for proficient resource management policies. In this paper, we investigate how to improve resource utilization and power consumption in cloud data centers when delivering services with statistically guaranteed Quality of Service (QoS). We assume that the service provider hosts different types of services, each of which has request classes with different QoS requirements. Different from the traditional approaches that distribute workloads with different QoS levels on different Virtual Machines (VMs), we introduce an approach to pack requests of the same service type, even with different QoS requirements, into the same VM, and to remove potential failure requests in time to improve resource usage and energy cost. We formally prove that our algorithm can statistically guarantee QoS conditions in terms of deadline miss ratios. We develop a cloud prototype to empirically validate our proposed methods and algorithm. Our experimental results demonstrate that our approach can significantly outperform other traditional approaches in terms of QoS guarantees, power consumption, resource demand and electricity cost. [ABSTRACT FROM PUBLISHER]

A new network structure called BCube Connected Crossbars (BCCC) was recently proposed. Its short diameter, good expandability and low cost make it a very promising topology for data center networks. However, it can utilize only two NIC ports of each server, which is suitable for nowadays technology, even though more NIC ports are available. Due to technology advances, servers with more NIC ports are emerging and they will become low-cost commodities some time later. In this paper, we propose a more general server-centric data center network structure, called GBC3, which can utilize inexpensive commodity off-the-shelf switches and servers with any fixed number of NIC ports and provide good network properties. Like BCCC, GBC3 has good expandability. When doing expansion, there is no need to alter the existing system but only to add new components into it. Thus the expansion cost that BCube suffers from can be significantly reduced in GBC3. We also introduce an addressing scheme and several efficient routing algorithms for one-to-one, one-to-all and one-to-many communications in GBC3 respectively. We make comprehensive comparisons between GBC3 and some popular existing structures in terms of several critical metrics, such as diameter, network size, bisection bandwidth and capital expenditure. We also conduct extensive experiments to evaluate GBC3, which show that GBC3 achieves the best flexibility to make tradeoff among all these critical metrics and it can suit for many different applications by fine tuning its parameters. [ABSTRACT FROM AUTHOR]

The computational and cooling power demands of enterprise servers are increasing at an unsustainable rate. Understanding the relationship between computational power, temperature, leakage, and cooling power is crucial to enable energy-efficient operation at the server and data center levels. This paper develops empirical models to estimate the contributions of static and dynamic power consumption in enterprise servers for a wide range of workloads, and analyzes the interactions between temperature, leakage, and cooling power for various workload allocation policies. We propose a cooling management policy that minimizes the server energy consumption by setting the optimum fan speed during runtime. Our experimental results on a presently shipping enterprise server demonstrate that including leakage awareness in workload and cooling management provides additional energy savings without any impact on performance. [ABSTRACT FROM AUTHOR]

Dealing with both dense and sparse data in parallel environments usually leads to two different approaches: To rely on a monolithic, hard-to-modify parallel library, or to code all data management details by hand. In this paper we propose a third approach, that delivers good performance while the underlying library structure remains modular and extensible. Our solution integrates dense and sparse data management using a common interface, that also decouples data representation, partitioning, and layout from the algorithmic and parallel strategy decisions of the programmer. Our experimental results in different parallel environments show that this new approach combines the flexibility obtained when the programmer handles all the details with a performance comparable to the use of a state-of-the-art, sparse matrix parallel library. [ABSTRACT FROM AUTHOR]

Reducing the carbon footprint of data centers is becoming a primary goal of large IT companies. Unlike traditional energy sources, renewable energy sources are usually intermittent and unpredictable. How to better utilize the green energy from these renewable sources in data centers is a challenging problem. In this paper, we exploit the opportunities offered by geographical load balancing, opportunistic scheduling of delay-tolerant workloads, and thermal storage management in data centers to facilitate green energy integration and reduce the cost of brown energy usage. Moreover, bandwidth cost variations between users and data centers are considered. Specifically, this problem is first formulated as a stochastic program, and then, an online control algorithm based on the Lyapunov optimization technique, called Stochastic Cost Minimization Algorithm (SCMA), is proposed to solve it. The algorithm can enable an explicit trade-off between cost saving and workload delay. Numerical results based on real-world traces illustrate the effectiveness of SCMA in practice. [ABSTRACT FROM PUBLISHER]

In this paper, we study the problem of achieving the optimal tradeoff between operation risk and expected energy cost for Internet data center (IDC) operators in deregulated electricity markets according to the risk preferences of IDC operators. To achieve the target above, we propose a risk-constrained stochastic programming decision framework. Then, we formulate a risk-constrained expected energy cost minimization problem with the uncertainties in spot price and workload. To solve the formulated problem, we use a decomposition-based cutting plane algorithm. Finally, extensive evaluations based on real-life data show the effectiveness of the proposed decision framework. [ABSTRACT FROM AUTHOR]

Transactional contention management policies show considerable variation in relative performance with changing workload characteristics. Consequently, incorporation of fixed-policy Transactional Memory (TM) in general purpose computing systems is suboptimal by design and renders such systems susceptible to pathologies. Of particular concern are Hardware TM (HTM) systems where traditional designs have hardwired policies in silicon. Adaptive HTMs hold promise, but pose major challenges in terms of design and verification costs. In this paper, we present the ZEBRA HTM design, which lays down a simple yet high-performance approach to implement adaptive contention management in hardware. Prior work in this area has associated contention with transactional code blocks. However, we discover that by associating contention with data (cache blocks) accessed by transactional code rather than the code block itself, we achieve a neat match in granularity with that of the cache coherence protocol. This leads to a design that is very simple and yet able to track closely or exceed the performance of the best performing policy for a given workload. ZEBRA, therefore, brings together the inherent benefits of traditional eager HTMs—parallel commits—and lazy HTMs—good optimistic concurrency without deadlock avoidance mechanisms—, combining them into a low-complexity design. [ABSTRACT FROM AUTHOR]

Cloud data center management is a key problem due to the numerous and heterogeneous strategies that can be applied, ranging from the VM placement to the federation with other clouds. Performance evaluation of cloud computing infrastructures is required to predict and quantify the cost-benefit of a strategy portfolio and the corresponding quality of service (QoS) experienced by users. Such analyses are not feasible by simulation or on-the-field experimentation, due to the great number of parameters that have to be investigated. In this paper, we present an analytical model, based on stochastic reward nets (SRNs), that is both scalable to model systems composed of thousands of resources and flexible to represent different policies and cloud-specific strategies. Several performance metrics are defined and evaluated to analyze the behavior of a cloud data center: utilization, availability, waiting time, and responsiveness. A resiliency analysis is also provided to take into account load bursts. Finally, a general approach is presented that, starting from the concept of system capacity, can help system managers to opportunely set the data center parameters under different working conditions. [ABSTRACT FROM AUTHOR]

This paper considers a stochastic optimization approach for job scheduling and server management in large-scale, geographically distributed data centers. Randomly arriving jobs are routed to a choice of servers. The number of active servers depends on server activation decisions that are updated at a slow time scale, and the service rates of the servers are controlled by power scaling decisions that are made at a faster time scale. We develop a two-time-scale decision strategy that offers provable power cost and delay guarantees. The performance and robustness of the approach is illustrated through simulations. [ABSTRACT FROM AUTHOR]

The cloud computing paradigm is attracting an increased number of complex applications to run in remote data centers. Many complex applications require parallel processing capabilities. Parallel applications of certain nature often show a decreasing utilization of CPU resources as parallelism grows, mainly because of the communication and synchronization among parallel processes. It is challenging but important for a data center to achieve a certain level of utilization of its nodes while maintaining the level of responsiveness of parallel jobs. Existing parallel scheduling mechanisms normally take responsiveness as the top priority and need nontrivial effort to make them work for data centers in the cloud era. In this paper, we propose a priority-based method to consolidate parallel workloads in the cloud. We leverage virtualization technologies to partition the computing capacity of each node into two tiers, the foreground virtual machine (VM) tier (with high CPU priority) and the background VM tier (with low CPU priority). We provide scheduling algorithms for parallel jobs to make efficient use of the two tier VMs to improve the responsiveness of these jobs. Our extensive experiments show that our parallel scheduling algorithm significantly outperforms commonly used algorithms such as extensible argonne scheduling system in a data center setting. The method is practical and effective for consolidating parallel workload in data centers. [ABSTRACT FROM AUTHOR]

Equal cost multiple path (ECMP) forwarding is the most prevalent multipath routing used in data center (DC) networks today. However, it fails to exploit increased path diversity that can be provided by traffic engineering techniques through the assignment of nonuniform link weights to optimize network resource usage. To this extent, constructing a routing algorithm that provides path diversity over nonuniform link weights (i.e., unequal cost links), simplicity in path discovery and optimality in minimizing maximum link utilization (MLU) is nontrivial. In this paper, we have implemented and evaluated the Penalizing Exponential Flow-spliTing (PEFT) algorithm in a cloud DC environment based on two dominant topologies, canonical and fat tree. In addition, we have proposed a new cloud DC topology which, with only a marginal modification of the current canonical tree DC architecture, can further reduce MLU and increase overall network capacity utilization through PEFT routing. [ABSTRACT FROM AUTHOR]

Enterprises typically operate multiple data center sites, each handling workloads according to an enterprise-level strategy. Sharing resources across multiple sites (or enterprises) brings up several important problems. Each site may have its own policies that govern its interactions with other remote sites. Different policies impact the system performance in different ways, and site administrators and system designers need to understand the effects of a given set of policies on potential workloads. In this paper, we describe an analysis methodology that determines the impact of policies on workloads, and we present results and validation for a prototypical multisite resource sharing system. Our analytical tool is capable of evaluating complex policies on a large-scale system and permits independent policies for each site so that policy makers can quickly evaluate several alternatives and their effects on the workloads before deploying them. [ABSTRACT FROM AUTHOR]

Data cube construction is a commonly used operation in data warehouses. Because of the volume of data that is stored and analyzed in a data warehouse and the amount of computation involved in data cube construction, it is natural to consider parallel machines for this operation. This paper addresses a number of algorithmic issues in parallel data cube construction. First, we present an aggregation tree for sequential (and parallel) data cube construction which has minimally bounded memory requirements. An aggregation tree is parameterized by the ordering of dimensions. We present a parallel algorithm based upon the aggregation tree. We analyze the interprocessor communication volume and construct a closed form expression for it. We prove that the same ordering of the dimensions in the aggregation tree minimizes both the computational and communication requirements. We also describe a method for partitioning the initial array and prove that it minimizes the communication volume. Finally, in the cases when memory may be a bottleneck, we describe how tiling can help scale sequential and parallel data cube construction. Experimental results from implementation of our algorithms on a cluster of workstations show the effectiveness of our algorithms and validate our theoretical results. [ABSTRACT FROM AUTHOR]

The bisection width of interconnection networks has always been important in parallel computing, since it bounds the speed at which information can be moved from one side of a network to another, i.e., the bisection bandwidth. Finding its exact value has proven to be challenging for some network families. For instance, the problem of finding the exact bisection width of the multidimensional torus was posed by Leighton [1, Problem 1.281] and has remained open for almost 20 years. We provide two general results that allow us to obtain upper and lower bounds on the bisection width of any product graph as a function of some properties of its factor graphs. The power of these results is shown by deriving the exact value of the bisection width of the torus, as well as of several $(d)$-dimensional classical parallel topologies that can be obtained by the application of the Cartesian product of graphs. We also apply these results to data centers, by obtaining bounds for the bisection bandwidth of the $(d)$-dimensional BCube network, a recently proposed topology for data centers. [ABSTRACT FROM AUTHOR]

Dynamic consolidation of virtual machines (VMs) is an effective way to improve the utilization of resources and energy efficiency in cloud data centers. Determining when it is best to reallocate VMs from an overloaded host is an aspect of dynamic VM consolidation that directly influences the resource utilization and quality of service (QoS) delivered by the system. The influence on the QoS is explained by the fact that server overloads cause resource shortages and performance degradation of applications. Current solutions to the problem of host overload detection are generally heuristic based, or rely on statistical analysis of historical data. The limitations of these approaches are that they lead to suboptimal results and do not allow explicit specification of a QoS goal. We propose a novel approach that for any known stationary workload and a given state configuration optimally solves the problem of host overload detection by maximizing the mean intermigration time under the specified QoS goal based on a Markov chain model. We heuristically adapt the algorithm to handle unknown nonstationary workloads using the Multisize Sliding Window workload estimation technique. Through simulations with workload traces from more than a thousand PlanetLab VMs, we show that our approach outperforms the best benchmark algorithm and provides approximately 88 percent of the performance of the optimal offline algorithm. [ABSTRACT FROM AUTHOR]

In today's data centers, precisely controlling server power consumption is an essential way to avoid system failures caused by power capacity overload or overheating due to increasingly high server density. While various power control strategies have been recently proposed, existing solutions are not scalable to control the power consumption of an entire large-scale data center, because these solutions are designed only for a single server or a rack enclosure. In a modern data center, however, power control needs to be enforced at three levels: rack enclosure, power distribution unit, and the entire data center, due to the physical and contractual power limits at each level. This paper presents SHIP, a highly scalable hierarchical power control architecture for large-scale data centers. SHIP is designed based on well-established control theory for analytical assurance of control accuracy and system stability. Empirical results on a physical testbed show that our control solution can provide precise power control, as well as power differentiations for optimized system performance and desired server priorities. In addition, our extensive simulation results based on a real trace file demonstrate the efficacy of our control solution in large-scale data centers composed of 5,415 servers. [ABSTRACT FROM AUTHOR]

Today's data centers face two critical challenges. First, various customers need to be assured by meeting their required service-level agreements such as response time and throughput. Second, server power consumption must be controlled in order to avoid failures caused by power capacity overload or system overheating due to increasing high server density. However, existing work controls power and application-level performance separately, and thus, cannot simultaneously provide explicit guarantees on both. In addition, as power and performance control strategies may come from different hardware/software vendors and coexist at different layers, it is more feasible to coordinate various strategies to achieve the desired control objectives than relying on a single centralized control strategy. This paper proposes Co-Con, a novel cluster-level control architecture that coordinates individual power and performance control loops for virtualized server clusters. To emulate the current practice in data centers, the power control loop changes hardware power states with no regard to the application-level performance. The performance control loop is then designed for each virtual machine to achieve the desired performance even when the system model varies significantly due to the impact of power control. Co-Con configures the two control loops rigorously, based on feedback control theory, for theoretically guaranteed control accuracy and system stability. Empirical results on a physical testbed demonstrate that Co-Con can simultaneously provide effective control on both application-level performance and underlying power consumption. [ABSTRACT FROM AUTHOR]

It has been widely acknowledged that online file hosting systems within the “cloud” of the Internet have provided valuable services to end users who wish to share files of any size. Such online hosting services are typically provided by dedicated servers, either in content distribution networks (CDNs) or large data centers. Server bandwidth costs, however, are prohibitive in these cases, especially when serving large volumes of files to a large number of users. Though it seems intuitive to take advantage of peer upload bandwidth to mitigate such server bandwidth costs in a complementary fashion, it is not trivial to design and fine-tune important aspects of such peer-assisted online hosting in a real-world large-scale deployment. This paper presents FS2You, a large-scale and real-world online file hosting system with peer assistance and semipersistent file availability. FS2You is designed to dramatically mitigate server bandwidth costs. In this paper, we show a number of key challenges involved in such a design objective, our architectural and protocol design in response to these challenges, as well as an extensive measurement study at a large scale to demonstrate the effectiveness of our design, using real-world traces that we have collected. To our knowledge, this paper represents the first attempt to design, implement, and evaluate a new peer-assisted semipersistent online file hosting system at a realistic scale. Since the launch of FS2You, it has quickly become one of the most popular online file hosting systems in mainland China, and a favorite in many online forums across the country. [ABSTRACT FROM PUBLISHER]

Causal consistency (CC) is an attractive consistency model for geo-replicated data stores because it hits a sweet spot in the ease-of-programming versus performance trade-off. We present a new approach for implementing CC in geo-replicated data stores, which we call Optimistic Causal Consistency (OCC). OCC's main design goal is to maximize data freshness. The optimism in our approach lies in the fact that the updates replicated to a remote data center are made visible immediately, without checking if their causal dependencies have been received. Servers perform the dependency check needed to enforce CC only upon serving a client operation, rather than on receipt of a replicated data item as in existing systems. OCC offers a significant gain in data freshness, which is of crucial importance for various types of applications, such as real-time systems. OCC's potentially blocking behavior makes it vulnerable to network partitions. We therefore propose a recovery mechanism that allows an OCC system to fall back on a pessimistic protocol to continue operating during network partitions. We implement POCC, the first causally consistent geo-replicated multi-master key-value data store designed to maximize data freshness. We show that POCC improves data freshness, while offering comparable or better performance than its pessimistic counterparts. [ABSTRACT FROM AUTHOR]

Realistic, relevant, and reproducible experiments often need input traces collected from real-world environments. In this work, we focus on traces of workflows—common in datacenters, clouds, and HPC infrastructures. We show that the state-of-the-art in using workflow-traces raises important issues: (1) the use of realistic traces is infrequent and (2) the use of realistic, open-access traces even more so. Alleviating these issues, we introduce the Workflow Trace Archive (WTA), an open-access archive of workflow traces from diverse computing infrastructures and tooling to parse, validate, and analyze traces. The WTA includes ${>}48$ > 48 million workflows captured from ${>}10$ > 10 computing infrastructures, representing a broad diversity of trace domains and characteristics. To emphasize the importance of trace diversity, we characterize the WTA contents and analyze in simulation the impact of trace diversity on experiment results. Our results indicate significant differences in characteristics, properties, and workflow structures between workload sources, domains, and fields. [ABSTRACT FROM AUTHOR]

Both power and performance are important concerns for enterprise data centers. While various management strategies have been developed to effectively reduce server power consumption by transitioning hardware components to lower power states, they cannot be directly applied to today's data centers that rely on virtualization technologies. Virtual machines running on the same physical server are correlated because the state transition of any hardware component will affect the application performance of all the virtual machines. As a result, reducing power solely based on the performance level of one virtual machine may cause another to violate its performance specification. This paper proposes PARTIC, a two-layer control architecture designed based on well-established control theory. The primary control loop adopts a multi-input multi-output control approach to maintain load balancing among all virtual machines so that they can have approximately the same performance level relative to their allowed peak values. The secondary performance control loop then manipulates CPU frequency for power efficiency based on the uniform performance level achieved by the primary loop. Empirical results demonstrate that PARTIC can effectively reduce server power consumption while achieving required application-level performance for virtualized enterprise servers. [ABSTRACT FROM AUTHOR]

Power control is becoming a key challenge for effectively operating a modern data center. In addition to reducing operating costs, precisely controlling power consumption is an essential way to avoid system failures caused by power capacity overload or overheating due to increasing high server density. Control-theoretic techniques have recently shown a lot of promise for power management because of their better control performance and theoretical guarantees on control accuracy and system stability. However, existing work oversimplifies the problem by controlling a single server independently from others. As a result, at the enclosure level where multiple high-density servers are correlated by common workloads and share common power supplies, power cannot be shared to improve application performance. In this paper, we propose an enclosure-level power controller that shifts power among servers based on their performance needs, while controlling the total power of the enclosure to be lower than a constraint. Our controller features a rigorous design based on an optimal Multi-Input-Multi-Output (MIMO) control theory. We present detailed control problem formulation and transformation to a standard constrained least-squares problem, as well as stability analysis in the face of significant workload variations. We then conduct extensive experiments on a physical testbed to compare our controller with three state-of-the-art controllers: a heuristic-based MIMO control solution, a Single-Input-Single-Output (SISO) control solution, and an improved SISO controller with simple power shifting among servers. Our empirical results demonstrate that our controller outperforms all the three baselines by having more accurate power control and up to 11.8 percent better benchmark performance. [ABSTRACT FROM AUTHOR]

In this study, we propose an energy-efficient database system called GreenDB running on clusters. GreenDB applies a workload-skewness strategy by managing hot nodes coupled with a set of cold nodes in a database cluster. GreenDB fetches popular data tables to hot nodes, aiming to keep cold nodes in the low-power mode in increased time periods. GreenDB is conducive to reducing the number of power-state transitions, thereby lowering energy-saving overhead. A prefetching model and an energy saving model are seamlessly integrated into GreenDB to facilitate the power management in database clusters. We quantitatively evaluate GreenDB's energy efficiency in terms of managing, fetching, and storing data. We compare GreenDB's prefetching strategy with the one implemented in Postgresql. Experimental results indicate that GreenDB conserves the energy consumption of the existing solution by up to 98.4 percent. The findings show that the energy efficiency of GreenDB can be optimized by tuning system parameters, including table size, hit rates, number of nodes, number of disks, and inter-arrival delays. [ABSTRACT FROM AUTHOR]

Data center networks demand high-performance, robust, and practical data plane load balancing protocols. Despite progress, existing work falls short of meeting these requirements. We design, analyze, and evaluate Luopan, a novel sampling based load balancing protocol that overcomes these challenges. Luopan operates at flowcell granularity similar to Presto. It periodically samples a few paths for each destination switch and directs flowcells to the least congested one. By being congestion-aware, Luopan improves flow completion time (FCT), and is more robust to topological asymmetries compared to Presto. The sampling approach simplifies the protocol and makes it much more scalable for implementation in large-scale networks compared to existing congestion-aware schemes. We provide analysis to show that Luopan's periodic sampling has the same asymptotic behavior as instantaneous sampling: taking 2 random samples provides exponential improvements over 1 sample. We conduct comprehensive packet-level simulations with production workloads. The results show that Luopan consistently outperforms state-of-the-art schemes in large-scale topologies. Compared to Presto, Luopan with 2 samples improves the 99.9%ile FCT of mice flows by up to 35 percent, and average FCT of medium and elephant flows by up to 30 percent. Luopan also performs significantly better than Local Sampling with large asymmetry. [ABSTRACT FROM AUTHOR]

Large-scale datacenters are facing increasing pressure of capping their carbon emission and power cost. Many leading-edge studies have started to explore server clusters running on multiple power sources. Existing approaches do not sufficiently consider the fine-grained power delivery to satisfy diverse requirements in datacenter, especially in the multi-tenant/colocation datacenter, which may yield low energy utilization. To address the emerging trend and new requirements, this article proposes a novel Datacenter inner Power Switch Network (DiPSN) to improve datacenter power efficiency and user satisfaction.DiPSNis a reconfigurable and easy-to-scale-out power architecture, which enables datacenter to distribute various power sources in a fine-grained manner. Moreover, a tailored machine learning based power source management framework is proposed forDiPSNto dynamically optimize user customized performance metrics and maximize datacenter revenue. Compared with conventional single-switch power distribution system, ourDiPSNcan be configured to improve solar energy utilization by 39.6 percent, reduce utility power cost by 11.1 percent and improve workload performance by 33.8 percent. Meanwhile, our design can extend battery lifetime by 9.3 percent. This work could provide valuable guidelines for designing heterogeneous power distribution architecture and management methodology in datacenters for improving user-customizable efficiency, sustainability and economy. [ABSTRACT FROM AUTHOR]