To achieve high-throughput and low-latency Remote Direct Memory Access (RDMA) communication in data center networks, load balancing is critical for preventing congestion and ensuring that traffic is efficiently distributed across available network paths. However, existing schemes may not effectively detect rerouting opportunities in continuous RDMA packet streams and may degrade in-order delivery, limiting their applicability to RDMA traffic. To address these limitations, we propose ProLet, a load balancing scheme that enables proactive probing and reroutes elephant flows at flowlet granularity in lossless RDMA networks. ProLet dynamically fine-tunes per-destination top-of-rack timeouts and enables effective in-network flowlet identification based on real-time network conditions. Meanwhile, it leverages lightweight mice flows as proactive probes to maintain network-wide congestion awareness. This allows ProLet to reroute elephant flows before congestion accumulates, mitigating the persistent queue buildup inherent in subflow-based schemes. Extensive numerical evaluations demonstrate that ProLet reduces average and tail flow completion time slowdowns by 69% and 79%, respectively, compared to state-of-the-art load balancing schemes.

To pursue better predictive performance in deep neural networks (DNNs), the size of learning models tends to increase, resulting in high computational requirements and training latency. In this paper, we investigate the problem of device placement to accelerate large-scale distributed DNN training. To address the challenge of finding an effective scheme due to its NP-hardness and the ever-increasing model size, we propose novel operator fusion and co-location schemes to reduce the search space while minimizing subsequent performance loss in training latency, enabling efficient device placement. We evaluate the performance of our design with real-world DNN benchmarks, and the results show that, compared to state-of-the-art approaches, our design achieves up to a 35 % reduction in DNN training latency and an order-of-magnitude improvement in placement search latency.

With the growing popularity of large-scale deep neural networks, efficient communication scheduling has become crucial in distributed deep learning systems to reduce overall training time. In multi-job distributed training scenarios, current communication scheduling methods do not effectively utilize the periodic communication patterns of deep learning training (DLT) jobs to reduce the potential link contention. When multiple tenants run concurrent jobs and compete for network resources, training performance can degrade due to increased network contention. In this paper, we focus on exploring the potential of leveraging periodic communication patterns in scheduling DLT jobs. We analyze the performance of static shift-based scheduling strategies based on the least common multiple (LCM) alignment in handling multi-job communication conflicts. Through theoretical analysis and validation via real-world experiments, we expose fundamental limitations of shift-based scheduling strategies, which fail to improve training throughput in about 73 % of multi-job scenarios. Our research work provides guidance for future research on understanding traffic patterns of DLT jobs and lays the groundwork for communication scheduling optimization in multi-tenant clusters.

Heterogeneous deep learning clusters commonly host a variety of distributed learning jobs. In such scenarios, the training efficiency of learning models is negatively affected by the slowest worker. To accelerate the training process, multiple learning jobs may compete for limited computational resources, posing significant challenges to multi-job placement among heterogeneous workers. This paper presents a heterogeneity-aware scheduler to solve the multi-job placement problem while taking into account job sizing and load balancing, minimizing the average Job Completion Time (JCT) of deep learning jobs. A novel scheme based on proportional training workload assignment, feasible solution categorization, and matching markets is proposed with theoretical guarantees. To further reduce the computational complexity for low latency decision-making and improve scheduling fairness, we propose to construct the sparsification of feasible solution categories through sampling, which has negligible performance loss in JCT. We evaluate the performance of our design with real-world deep neural network benchmarks on heterogeneous computing clusters. Experimental results show that, compared to existing solutions, the proposed sampling-based scheme can achieve 1) results within 2.04% of the optimal JCT with orders-of-magnitude improvements in algorithm running time, and 2) high scheduling fairness among learning jobs.

The rapid growth of data processing demands in large-scale data centers (DCs) has led to increased brown energy (BE) consumption, which has negative environmental impacts. Since most DCs are now powered by both BE and renewable energy (RE), migrating workloads from DCs with insufficient RE to DCs with sufficient RE can decrease the total BE consumption in the network. However, selecting a destination DC is challenging due to the uncertainty of the network and the additional cost associated with using network devices for the migration. This paper proposes to optimize the DC selection and the efficient virtual machine transfer between DCs, minimizing the costs of BE consumption, optical network devices, and migration. Specifically, we formulate the DC selection as a multi-armed bandit problem and estimate the lowest migration cost at each round using the lower confidence bound. We adopt the optical grooming technique to reduce the cost of optical devices used during the migration. We compare our algorithm with the KUBE and -Greedy algorithms on the NSFNET and show that it reduces the total cost by 4.6% and 12.8%, respectively, while having lower regret. We demonstrated the effectiveness of optical grooming by achieving a 12 % reduction in network costs.

Recent works on long music audio inpainting have focused on unconditionally generating new segments to inpaint corrupted audio segments. However, the information about these segments may differ significantly from the original. To solve this problem, we propose MAID (Music Audio Inpainting DDPM), a model for music audio inpainting based on DDPM (Denoising Diffusion Probability Model). The model is capable of unconditional and conditional inpainting of music audio: (a) in the unconditional inpainting task, MAID is capable of inpainting gaps with a length between 200 ms and 1600 ms; (b) in the conditional inpainting task, the model can generate new segments similar to the original segments based on the piano-rolls corresponding to the gaps. Experiments show that MAID performs better than the baseline.

In this paper, we study the open problems raised by Karp et al. in FOCS 2000, where the authors formulated the end-to-end congestion control as a repeated game between a flow and an adversary. They mentioned several open problems including finding equilibria in a more realistic game model for the situation where the available bandwidth is a result of competition among multiple flows instead of being chosen by an adversary, and designing the randomized algorithm to deal with the dynamic change of network bandwidth. Although there have been many game-theoretic works for congestion control, to the best of our knowledge, the above two problems still remain unsolved over the past decades. We take a step further to address the above two problems by first modeling the end-to-end congestion control as a repeated unknown general-sum game among multiple flows with bandit feedback. Each flow is a player in this unknown game, making decisions on how many packets to send. The throughput for each flow depends on all the flows' rates and the network capacity. The unknown setting and bandit feedback capture the essence of end-to-end congestion control: each flow has no information about others (e.g., the number, actions, and packet loss of other flows), and only receives limited information for its chosen action. Then, we propose a randomized no-regret learning algorithm for each flow called LUC based on a swap-regret-minimizing technique. We prove that LUC can guarantee a polynomial-time convergence rate to correlated equilibria in the multi-player setting. Finally, we have implemented LUC through the Linux kernel, and conducted extensive fairness-related experiments in Mininet and trace-driven experiments with Pantheon to show that each flow with LUC can fairly share the bandwidth in homogeneous scenarios, and be competitive but TCP-friendly in heterogeneous scenarios.

Erasure codes have been widely used to enhance data resiliency with low storage overheads. However, in geo-distributed cloud storage systems, erasure codes may incur high service latency as they require end users to access remote storage nodes to retrieve data. An elegant solution to achieving low latency is to deploy caching services at the edge servers close to end users. In this paper, we propose adaptive and scalable caching schemes to achieve low latency in the cloud-edge storage system. Based on the measured data popularity and network latencies in real time, an adaptive content replacement scheme is proposed to update caching decisions upon the arrival of requests. Theoretical analysis shows that the reduced data access latency of the replacement scheme is at least 50% of the maximum reducible latency. With the low computation complexity of our design, nearly no extra overheads will be introduced when handling intensive data flows. For further performance improvements without sacrificing its efficiency, an adaptive content adjustment scheme is presented to replace the subset of cached contents that incur the aforementioned performance loss. Driven by real-world data traces, extensive experiments based on Amazon Simple Storage Service demonstrate the effectiveness and efficiency of our design.

Caching has been considered as a promising solution to achieve low latency in distributed erasure coded storage systems. The previous research work categorizes all feasible caching decisions into a set of cache partitions, and then obtains the optimal solution by applying the market clearing price on each cache partition. While enjoying the ultimate performance of low data access latency, the optimal scheme suffers from high computation overheads when applied to large-scale storage systems. This paper presents SampleX, which constructs the sparsification of cache partitions through sampling to approximate the optimal caching scheme with substantially reduced computation complexity. Theoretical analysis guarantees the performance of SampleX. Furthermore, SampleX is implemented in a streaming fashion, capturing the characteristics of recent traffic for online cache content replacement. Trace-driven experimental results show that online SampleX is up to 95× faster than the state-of-the-art online scheme while only incurring a performance loss of 0.81%.

Machine reading comprehension (MRC) is a chal-lenging task, which has the long-standing goal of developing models that can determine the answer based on the given passage and question. Recent works have achieved promising results; however, most of them lose sight of the multiplicate information carried by the question. Inspired by the hot issue of research in the educational field, we propose a novel machine pre-reading activity over questions considering an efficient reading comprehension process of humans. Concretely, in the proposed pre-reading activity, the machine first pre-reads the question text and distinguishes the search term of the question using a question-answer type extraction algorithm (QAE). Second, the MRC model leverages two new search and enhancement methods to point out the most informative contents contained in the question, and highlight the question-based information in the passage accordingly. Third, armed with valid information, we introduce a question-based feature shunt module before the multi-head predictor. Moreover, we propose a latent A/B answer type, which not only makes the feature shunt more accurate but also enlarges the range of answer types. Experiments show that our strategies and their combinations achieve considerable improvements compared with existing methods.

Low-latency data service is an increasingly critical challenge for data center applications. In modern distributed storage systems, proper data placement helps reduce the data movement delay, which can contribute to the service latency reduction tremendously. Existing data placement solutions have often assumed the prior distribution of data requests or discovered it via trace analysis. However, data placement is a difficult online decision-making problem faced with dynamic network conditions and time-varying user request patterns. The conventional static model-based solutions are less effective to handle the dynamic system. With an overall consideration of data movement and analytical latency, we develop a reinforcement learning-based framework DataBot+, automatically learning the optimal placement policies. DataBot+ adopts neural networks, trained with a variant of$Q$-learning, whose input is the real-time data flow measurements and whose output is a value function estimating the near-future latency. For instantaneous decision making, DataBot+ is decoupled into two asynchronous production and training components, ensuring that the training delay will not introduce extra overheads to handle the data flows. Evaluation results driven by real-world traces demonstrate the effectiveness of our design.

Erasure codes have been widely considered as a promising solution to enhance data reliability at low storage costs. However, in modern geo-distributed storage systems, erasure codes may incur high data access latency as they require data retrieval from multiple remote storage nodes. This hinders the extensive application of erasure codes to data-intensive applications. This paper proposes novel caching schemes to achieve low latency in distributed coded storage systems. Assuming that future data popularity and network latency information are available, an offline caching scheme is proposed to explore the optimal caching solution for low latency. The proposed scheme categorizes all feasible caching decisions into a set of cache partitions, and then obtains the optimal caching decision through market clearing price for each cache partition. Furthermore, guided by the optimal scheme, an online caching scheme is proposed according to the measured data popularity and network latency information in real time, without the need to completely override the existing caching decisions. Both theoretical analysis and experiment results demonstrate that the online scheme can approximate the offline optimal scheme well with dramatically reduced computation complexity.

The electrocardiogram sensor network (ECG-SN) is a medical monitoring system based on IoT technology, which can detect heart bioelectric signals in real time. But the ECG signal is vulnerable to human mobility and sensitive to the Age of Information (AoI). In this article, we first analyze the impact of human mobility on channel fading based on a real-world activity trace data set and design a two-state ECG work model based on the tradeoff between the ECG signal quality and energy consumption. Furthermore, an AoI model is proposed to evaluate the data timeliness. Then, an online transmission optimization algorithm is proposed to maximize the system utility by optimizing the sampling rate, transmission power, and data dropping rate. Furthermore, performance analysis presents the bounds for data buffer, battery capacity, and AoI. Numerical results show the dynamics of the system and the impact of the parameters on system performance, which verify that the proposed design has a larger utility and a smaller AoI in comparison with two benchmark schemes.

Infrastructure-as-a-Service clouds in geo-distributed data centers offer various pricing options, including on-demand and reserved instances, which provide an elastic and cost-effective infrastructure to support High Performance Computing (HPC) applications. In this paper, we propose an instance reservation based cloud service framework, modeling the cost-minimizing reservation decision issue as an NP-hard integer programming problem for distributed data centers. To ease its computation complexity, two algorithms are proposed to minimize the HPC service cost with the worst-case performance guarantees: an offline heuristic-greedy algorithm, and a rolling-horizon based online algorithm when only short-term demand prediction is available. Facing fluctuating demands, instance reservation in a single data center may incur the highly underutilized capacity. To address this issue for further cost reduction, we extend the scheme with a novel cloud broker federation based resource sharing mechanism, reallocating already reserved but unused instances to computation-intensive and short-lived tasks for continuous execution without interruption. Extensive evaluations driven by large-scale trace-based datasets demonstrate that the proposed mechanism can effectively handle large volumes of service requests, saving considerable service costs with higher reservation resource utilization.

Mobile edge computing (MEC) has been considered as a promising technology to handle computation-intensive and delay-sensitive tasks in the Internet of Things (IoT) ecosystem, such as smart city and smart tourism. However, due to user mobility, edge servers with fixed deployment are not flexible enough to handle time-varying user tasks in hot-spot areas. In this article, a novel online unmanned aerial vehicle (UAV)-mounted edge server dispatching scheme is proposed to provide flexible mobile-to-MEC services. UAVs are dispatched to the appropriate hover locations by geographically merging tasks into several hot-spot areas. Theoretical analysis guarantees the worst case performance bound. Extensive evaluation driven by real-world mobile requests shows that while maintaining a good latency fairness, the mobile server dispatching scheme can serve more user equipments (UEs) as well as achieve a high resource utilization. Moreover, the hybrid scheme can satisfy even more user demands while dispatching fewer UAVs with a higher server utilization.

In geo-distributed cloud storage systems, data replication has been widely used to serve the ever more users around the world for high data reliability and availability. How to optimize the data replica placement has become one of the fundamental problems to reduce the inter-node traffic and the system overhead of accessing associated data items. In the big data era, traditional solutions may face the challenges of long running time and large overheads to handle the increasing scale of data items with time-varying user requests. Therefore, novel offline community discovery and online community adjustment schemes are proposed to solve the replica placement problem in a scalable and adaptive way. The offline scheme can find a replica placement solution based on the average read/write rates for a certain period of time. The scalability can be achieved as 1) the computation complexity is linear to the amount of data items and 2) the data-node communities can evolve in parallel for a distributed replica placement. Furthermore, the online scheme is adaptive to handle the bursty data requests, without the need to completely override the existing replica placement. Driven by real-world data traces, extensive performance evaluations demonstrate the effectiveness of our design to handle large-scale datasets.

Mobile edge computing (MEC) offloads computation-intensive applications and overcomes the long latency by pushing data traffic towards the network edges. With base stations (BSs) densely deployed in a hot-spot area to improve user experience, mobile user equipments (UEs) have multiple choices to offload tasks to edge servers by jointly considering both the channel condition and the computing capacity. However, precise full system information is hard to be synchronized between BSs and UEs for mobility management decision making. In this paper, a Q-Iearning based mobility management scheme is proposed to handle the system information uncertainties. Each UE observes the task delay as an experience and automatically learns the optimal mobility management strategy through trial and error. Simulations show that the proposed scheme manifests the superiority in dealing with the uncertainties. Compared with the traditional received signal strength-based handover scheme, the proposed scheme reduces the task delay by about 30%.

In this paper, we propose a novel real-time sound event detection framework, which combines multi-label learning and edge computing, to classify and localize abnormal sound events for city surveillance. Multiple devices equipped with acoustic sensors are deployed to collect the audio information. A learning-based approach is introduced to address the difficulties of accurately classifying the temporally overlapping acoustic events in a noisy environment. Then, edge computing is adopted to handle the high processing complexity of the learned analytics. Computation-intensive tasks of classification and localization can be offloaded to the nearby edge server for low-latency sound detection. An ensemble-based cooperative decision-making algorithm is also presented to aggregate the information from distributed devices in order to obtain better classification results. Extensive evaluations show the effectiveness of edge computing which helps reduce the time latency as well as the superiority of cooperative post-processing on the edge server to obtain a high accuracy.

Low-latency data access is an important challenge for data center networks. Proper placement of the data items can reduce the data travel time in the distributed storage systems, which contributes significantly to the latency reduction. Most existing data placement approaches have often assumed the prior distribution of data requests or discovered so through trace analysis. However, the traditional static model-based solutions are less effective to handle the system uncertainties in a dynamic environment. We present DataBot, a reinforcement learning-based adaptive framework, to learn the optimal data placement policies faced with the dynamic network conditions and time-varying request patterns. DataBot utilizes a neural network, trained with a variant of Q-learning, whose input is the realtime data flow measurements and whose output is a value function estimating the near-future latency. For rapid decision making, DataBot is divided into two decoupled production and training components, ensuring that the convergence time of the training will not introduce more overheads to serve the read/write requests. Evaluation results demonstrate that the average write and read latency of the whole system can be lowered by about 35% and 40%, respectively.

Recently, mobile cloud offloading is a promising technique to deal with the increasingly complex applications on mobile devices, meeting the ever- increasing energy requirements. However, cloud offloading with multiple mobile devices may cause considerable mutual interference, which may result in intolerable time delay and more energy consumption. In this paper, a novel offloading decision method is investigated to minimize the total energy consumption of mobile devices over cellular networks. Generally, mobile devices can execute a sequence of tasks in parallel with different characteristics, i.e., communication- intensive and computation-intensive. And recent advances show that only computation-intensive tasks are applicable to be offloaded for energy saving. The offloading decision issue is formulated as a NP- hard combinatorial optimization problem with the time deadline and communication quality constraints. Combining the problem linearization method and decision variables mapping from integer to the real domain, a rapid and efficient iterative approximation method is proposed, helping the cloud controller to select the best tasks for offloading aiming at minimizing the total energy consumption. Numerical simulation demonstrates that considerable energy can be saved with the proposed task offloading method in mobile cloud scenarios.

In cloud market, cloud providers offer diverse service options, including on-demand instances and reserved instances. Generally reserved instance price is cheaper than on-demand instance, but excessive reservation may result in high capacity underutilization. To improve resource utilization rate and reduce providers' service cost, a steady broker federation is necessary to propose. Broker federation coordinate cloud providers service demands in geo-distributed data centers through deciding when and how many instances to reserve. Firstly, the service cost optimization problem is formulated as a nonlinear integer programming model. Then a hybrid algorithm combining ant colony with particle swarm optimization is presented to reduce computational complexity and providers service cost. Extensive simulations driven by large-scale Parallel Workloads Archive demonstrate the effectiveness and efficiency of the hybrid algorithm.

This paper presents a new DC-DC converter that is designed to provided selected polarities at desired voltage levels. With the new switching converter topology, the converter can boost and buck the input voltage and theoretically achieve infinite levels for either positive or negative polarities. This converter can provide a wide range of output voltage and use only one switch to perform the power conversion which simplifies the design of the gate drive. This configuration can basically satisfy the need of those applications which require voltage polarity selection Simulations results demonstrate the operation of the proposed circuit. Experimental data validate the simulation and effectiveness of the polarity and amplitude generation. Circuit mathematical models and state variables are demonstrated in each voltage polarity.

In cloud computing, Infrastructure-as-a-Service (IaaS) cloud providers can offer two types of purchasing plans for cloud users, including on-demand plan and reservation plan. Generally reservation price is cheaper than on-demand price, while reservation plan may cause highly underutilized capacity problem. How to joint optimize the service cost and the resource utilization for clouds is a critical issue. To address this issue, a novel steady broker federation is developed to coordinate service demands in this paper. And the optimal reservation problem can be formulated as a nonlinear integer programming model. Then a fine-grained heuristic algorithm is proposed to reduce its computational complexity and obtain quasi-optimal solutions. Numerical simulations driven by large-scale Parallel Workloads Archive demonstrate that the proposed approach can save considerable costs for cloud users and improves the resource utilization for IaaS cloud providers.

Although several XML query-processing techniques have been proposed to do structural joins, no work has been done to handle aggregate structural join, which computes the aggregate value for two given XML element sets. One important kind of aggregate structural join is to determine the exact selectivity of a structural join, which proves to be crucial for overall path expression optimization and on-line systems. Related work has been done to estimate the selectivity of a path expression by applying summary techniques for XML data based on either tree simplification or a Markov model. However, because of either space restrictions or performance issues (for example, previous approaches may need to store the whole XML tree to answer an aggregate structural join), they are not applicable for aggregate structural joins. In this paper, we consider a modified aggregate B-tree, which we call a XA-tree, to index the aggregate attribute values of XML elements. Since it is based on the B-tree, the XA-tree supports query and update efficiently and places little additional burden on database administrators. Two basic algorithms are proposed to compute aggregate structural joins efficiently by utilizing the XA-tree. Furthermore, a heuristic is also proposed to further improve the performance significantly. Extensive experiments have confirmed that our approaches greatly outperform competitors by an order of magnitude.