In cloud computing, a provider leases its computing resources in the form of virtual machines to users, and price is charged for period they are used. Though static pricing dominant strategy today's market, intuitively ought be dynamically updated improve revenue. The fundamental challenge design an optimal dynamic policy, with presence stochastic demand perishable resources, so that expected long-term revenue maximized. this paper, we make three contributions addressing question. First,...

Software defined networking is becoming increasingly prevalent in data center networks for its programmability that enables centralized network configuration and management. However, since switches are statically assigned to controllers, traffic dynamics cause load imbalance among the controllers. As a result, some controllers not fully utilized, while connected overloaded may experience long response times. In this paper, we consider dynamic controller assignment so as minimize average time...

With the evolution of Network Function Virtual-izaiton (NFV), enterprises are increasingly outsourcing their network functions to cloud. However, using virtualized (VNFs) provide flexible services in today's cloud is challenging due inherent difficulty intelligently scaling VNFs cope with traffic fluctuations. To best utilize resources, NFV providers need dynamically scale VNF deployments and reroute demands for customers. Since most existing work reactive nature, we seek a proactive...

Various datacenter network (DCN) load balancing schemes have been proposed in the past decade. Unfortunately, most of these solutions designed for lossy DCNs do not work well Priority Flow Control (PFC) enabled lossless DCNs, primarily due to reason that individual congestion signals used solutions, e.g., link load, queue length, Round Trip Time (RTT) and Explicit Congestion Notification (ECN), may be able correctly or timely reflect hop-by-hop PFC pausing. This paper first reveals above...

There has been a recent trend that video-on-demand (VoD) providers such as Netflix are leveraging resources from cloud services for multimedia streaming. In this paper, we consider the scenario VoD provider can make reservations bandwidth guarantees service to guarantee streaming performance in each video channel. We propose predictive resource auto-scaling system dynamically books minimum multiple data centers match its short-term demand projections. exploit anti-correlation between demands...

Many cloud services are running on geographically distributed datacenters for better reliability and performance. We consider the emerging problem of joint request mapping response routing with in this paper. formulate as a general workload management optimization. A utility function is used to capture various performance goals, location diversity electricity bandwidth costs realistically modeled. To solve large-scale optimization, we develop algorithm based alternating direction method...

Software defined networking is increasingly prevalent in data center networks for it enables centralized network configuration and management. However, since switches are statically assigned to controllers provisioned, traffic dynamics may cause long response time incur high maintenance cost. To address these issues, we formulate the dynamic controller assignment problem (DCAP) as an online optimization minimize total cost caused by on cluster of controllers. By applying randomized fixed...

Short TCP flows that are critical for many interactive applications in data centers plagued by large and head-of-line blocking switches. Hash-based load balancing schemes such as ECMP aggravate the matter result long-tailed flow completion times (FCT). Previous work on reducing FCT usually requires custom switch hardware and/or protocol changes. We propose RepFlow, a simple yet practically effective approach replicates each short to reduce times, without any change switches or host kernels....

Multilabel image classification aims to assign images multiple possible labels. In this task, each may be associated with labels, making it more challenging than the single-label problems. For instance, convolutional neural networks (CNNs) have not met performance requirement in utilizing statistical dependencies between labels study. Additionally, data imbalance is a common problem machine learning that needs considered for multilabel medical classification. Furthermore, concatenation of...

Dynamic spectrum trading amongst small cognitive users is fundamentally different along two axes: temporal variation, and spatial variation of user demand channel condition. We advocate that a secondary market, analogous to the stock be established for dynamically trade among themselves their holdings obtained in primary market from legacy owners. design mechanism based on dynamic double auctions, creating marketplace air match bandwidth with supply. In analysis we prove important economic...

Datacenters consume an enormous amount of energy with significant financial and environmental costs. For geo-distributed datacenters, a workload management approach that routes user requests to locations cheaper cleaner electricity has been shown be promising lately. We consider two key aspects have not explored in this approach. First, through empirical studies, we find the efficiency cooling system depends directly on ambient temperature, which exhibits degree geographical diversity....

We present Anchor, a general resource management architecture that uses the stable matching framework to decouple policies from mechanisms when mapping virtual machines physical servers. In clients and operators are able express variety of distinct as they deem fit, these captured preferences in framework. The highlight Anchor is new many-to-one theory efficiently matches VMs with heterogeneous needs servers, using both offline online algorithms. Our theoretical analyses show convergence...

Recently, data center carbon emission has become an emerging concern for the cloud service providers. Previous works are limited on cutting down power consumption of centers to defuse such a concern. In this paper, we show how spatial and temporal variabilities electricity footprint can be fully exploited further green running top geographically distributed centers. We jointly consider cost, level agreement (SLA) requirement, reduction budget. To navigate three-way tradeoff, take advantage...

Data centers consume a large amount of energy and incur substantial electricity cost. In this paper, we study the familiar problem reducing data center cost with two new perspectives. First, find, through an empirical contracts from electric utilities powering Google centers, that demand charge per kW for maximum power used is major component total Second, many services such as web search tolerate partial execution requests because response quality concave function processing time. Microsoft...

Recently, datacenter carbon emission has become an emerging concern for the cloud service providers. Previous works are limited on cutting down power consumption of datacenters to defuse such a concern. In this paper, we show how spatial and temporal variabilities electricity footprint can be fully exploited further green running top geographically distributed datacenters. Specifically, first verify that cost minimization conflicts with minimization, based empirical study several...

We study the infinite horizon dynamic pricing problem for an infrastructure cloud provider in emerging computing paradigm. The provider, such as Amazon, provides capacity form of virtual instances and charges customers a time-varying price period they use instances. provider's is then to find optimal policy, face stochastic demand arrivals departures, so that average expected revenue maximized long run. adopt management framework tackle problem. Optimality conditions structural results are...

Offchain networks emerge as a promising solution to address the scalability challenge of blockchain. Participants make payments through offchain instead committing transactions on-chain. Routing is critical performance networks. Existing solutions use either static routing with poor or dynamic high overhead obtain channel balance information. In this paper, we propose Flash, new that leverages unique characteristics in strike better tradeoff between path optimality and probing overhead. By...

We study the resource allocation problem in an OFDMA-based cooperative cognitive radio network, where secondary users relay data for primary order to gain access spectrum. In light of user and channel diversity, we first propose FLEC, a novel flexible cooperation scheme. It allows freely optimize use channels transmitting along with their own, maximize performance. Further, formulate unifying optimization framework based on Nash bargaining solutions fairly efficiently allocate resources...

Recently, a cooperative paradigm for single-channel cognitive radio networks has been advocated, where primary users can leverage secondary to relay their traffic. However, it is not clear how such cooperation be exploited in multi-channel effectively. Conventional entails that data on one channel relayed exactly the same channel, which inefficient with and user diversity. Moreover, selfishness of complicates critical resource allocation problem, as both parties target at maximizing own...

We present a study of pricing cloud resources in this position paper. Our objective is to explore and understand the interplay between economics systems designs proposed by recent research. develop general model that captures resource needs various applications usage computing. show uniform price does not suffer any revenue loss compared first-order discrimination. then consider alternative strategies provider can use improve revenue, including throttling performance guarantees, enabled...

To maximize data center network utilization, the SDN control plane needs to frequently update as conditions change. Since each switch updates its flow table independently and asynchronously, state transition -- if done directly from initial final stage may result in serious flash congestion packet loss. Prior work strives find a congestion-free plan with multiple stages, property that there will be no independent of order. Yet requires part link capacity left vacant decreases utilization...

Middleboxes are widely deployed to perform various network functions ensure security and improve performance. The recent trend of Network Function Virtualization (NFV) makes it easy for operators deploy software implementations these on commodity servers. However, virtual consume different amounts resources when processing packets. Thus a multi-resource load balancing (MRLB) mechanism is needed efficiently utilize server resources. MRLB problem in the context NFV fundamentally from...

Medical image segmentation is a fundamental step in medical analysis and diagnosis. In recent years, deep learning networks have been used for precise segmentation. Numerous improved encoder-decoder structures proposed various tasks. However, high-level features gained more research attention than the abundant low-level early stages of Consequently, edge feature maps has limited, which can lead to ambiguous boundaries predicted results. Inspired by network mechanism, this study investigates...