With the increasing time drivers spent behind the wheel, availability of accurate, reliable, and comprehensive information on real-time traffic conditions is becoming increasingly important for individual road users, ...
详细信息
ISBN:
(纸本)9798350328127
With the increasing time drivers spent behind the wheel, availability of accurate, reliable, and comprehensive information on real-time traffic conditions is becoming increasingly important for individual road users, industry, and traffic authorities. Due to a growing amount of traffic combined with insufficient transportation infrastructure, traffic congestion is one of the major challenges modern society has to face in urban areas. While this results in serious environmental issues such as pollution, congestions also have significant effects on drivers' stress levels, which further is an essential factor in driver behavior and road safety. In this paper, we analyze real-time traffic situations on a fine-granular level by computing the local congestion factor, which represents traffic conditions on currently traversed route segments. We show that travel time predictions for short- and mid-term routes provided by four major public traffic information providers can be used for computing congestion factors in real-time, which can then be used as input for driver-centric applications.
New vehicular applications demand more computing power and real-time processing. As modern vehicles are equipped with computationally powerful but often redundant and under-utilized onboard units for autonomous drivin...
详细信息
ISBN:
(纸本)9798350328127
New vehicular applications demand more computing power and real-time processing. As modern vehicles are equipped with computationally powerful but often redundant and under-utilized onboard units for autonomous driving, a network of connected vehicles can form a vehicular cloud (VC) that can provide computing services among themselves or to other devices. In this paper, we evaluate the computing performance of a VC on a highway in congested traffic. We assume that the vehicles join and leave VC at random times. Thus, the number of vehicles in the VC will be time-varying. The residency times of the vehicles in the VC will be correlated because of traffic congestion. To enable the realization of VC in the future, we need to know its computing performance that considers its dynamic nature and concurrent execution of the tasks. In this work, we determine the completion time of a job with multiple tasks with random execution times. More specifically, we derive the probability density function of the job completion time as a function of the system parameters. We provide numerical results to demonstrate the utilization of the analysis and simulation results to confirm the correctness of the analysis.
Recently, ZNS SSDs have been actively researched to handle the functions of the FTL directly on the host system. ZNS SSDs can improve the I/O performance and spatial efficiency of SSDs by eliminating GC and eliminatin...
详细信息
ISBN:
(纸本)9798350339864
Recently, ZNS SSDs have been actively researched to handle the functions of the FTL directly on the host system. ZNS SSDs can improve the I/O performance and spatial efficiency of SSDs by eliminating GC and eliminating overprovisioning. In this paper, we analyze the performance by running distributed applications on a file system which supports ZNS SSDs. We found that much of the time difference occurs in the process of executing open and unlink operations rather than the performance of read and write operations. It has been confirmed that the performance difference of read/write operations is not significant for the applications. If structural optimization is made on file system, it has been found that ZNS SSDs are better than CNS SSDs of the same capacity in terms of price.
Convolutional Neural Networks (CNNs) are an essential part of distributed neuromorphic systems and require an efficient strategy for deployment. We proposed a synergetic distributed mapping mechanism for CNNs to addre...
详细信息
ISBN:
(纸本)9798350380903;9798350380910
Convolutional Neural Networks (CNNs) are an essential part of distributed neuromorphic systems and require an efficient strategy for deployment. We proposed a synergetic distributed mapping mechanism for CNNs to address the problem of parallel computing and resource optimization of CNNs in neuromorphic systems. A dataset recording soccer positions containing 4000 grayscale images of 105*105 pixels was built. A CNN model was constructed based on it and deployed in the developed neuromorphic system BrainS. As tested in the robotic platform, the model has an accuracy of 95.2% in the test set and can accurately control the movement of the omnidirectional robot. The Root Mean Square Error (NMSE) between BrainS and GPU for all layers during the forward computation of the model was 2.00 x 10(-6). By adjusting the number of guard bits in the computing process, the system's time consumption can be reduced. The proposed method in this paper provides a practical and effective solution for CNN deployment in neuromorphic systems while guaranteeing accuracy.
In distributedcomputing, coding techniques are shown to be an effective solution for mitigating the impact of stragglers. However, previous research on coding has predominantly focused on homogeneous worker environme...
详细信息
ISBN:
(纸本)9798350385304;9798350385298
In distributedcomputing, coding techniques are shown to be an effective solution for mitigating the impact of stragglers. However, previous research on coding has predominantly focused on homogeneous worker environments, overlooking the fact that real-world systems often consist of heterogeneous workers with varying computing and communication capabilities. Specifically, uniform load allocation, without considering worker heterogeneity, can result in substantial latency losses. In this paper, we propose a load allocation strategy designed for distributedsystems with group heterogeneity, where workers in the same group have similar computing and communication capabilities, but workers of different groups do not. By exploring group-wise MDS codes, we determine the optimal code dimension and optimal computation loads for individual groups. Our proposed approach demonstrates comparable computation times to existing methods, while exhibiting the advantage of much shorter encoding and decoding times.
Network policy plays a crucial role in cloud-native networking, especially in multi-tenant scenarios. It provides precise control over connectivity by specifying source and destination endpoints, traffic types, and ot...
详细信息
ISBN:
(纸本)9798350386066;9798350386059
Network policy plays a crucial role in cloud-native networking, especially in multi-tenant scenarios. It provides precise control over connectivity by specifying source and destination endpoints, traffic types, and other criteria to allow or deny traffic. However, manual configuration of these policies introduces the risk of errors, leading to isolation violations or network service unavailability. Therefore, network policy verification is essential for maintaining security and quality of service in cloud-native networking. Currently, a naive approach involves individually checking each policy within the cluster, which can take over 100s for verification in a cluster size of over 100k. Existing verification frameworks, like Kano and Verikube, improve performance by leveraging pre-filtering and Satisfiability Modulo Theories (SMT) solvers, achieving a 3.12x to 12.99x performance boost over the naive baseline. However, as network policy changes rapidly within 100ms in real cloud-native networks, both frameworks need over 10s to perform verification for cluster sizes over 100k, which is far from satisfying. To overcome these issues, we propose and implement a novel network policy verification framework NPV, which utilizes the policy-label pre-filter process with bitwise compression. We further enhance the policy verification algorithm with a policy-namespace divide-and-conquer strategy to improve the data-level parallelism. We implement NPV on commodity servers and evaluate its performance using real network policy datasets. Our experiments indicate that, compared with the state-of-the-art methods, NPV can achieve up to 139.00x to 651.06x improvement in verification time compared to Kano and Verikube, with 65% less memory usage.
Machine Learning as a Service (MLaaS) has paved the way for numerous applications for resource-limited clients, such as IoT/mobile users. However, it raises a great challenge for privacy, including both the data priva...
详细信息
ISBN:
(纸本)9798350386066;9798350386059
Machine Learning as a Service (MLaaS) has paved the way for numerous applications for resource-limited clients, such as IoT/mobile users. However, it raises a great challenge for privacy, including both the data privacy of clients and model privacy of the server. While there have been extensive studies on privacy-preserving MLaaS, a direct adoption of current frameworks leads to intractable efficiency bottleneck for MLaaS with resource constrained clients. In this paper, we focus on MLaaS with resource constrained clients and propose a novel privacy-preserving framework called SPOT to address a unique challenge, the memory constraint of such clients, such as IoT/mobile devices, which results in significant computation stalls at the server in privacy-preserving MLaaS. We develop 1) a novel structure patching scheme to enable independent computations for sequential inputs at the server to eliminate the computation stall, and 2) a patch overlap tweaking scheme to minimize overlapped data between adjacent patches and thus enable more efficient computation with flexible cryptographic parameters. SPOT demonstrates significant improvement on computation efficiency for MLaaS with IoT/mobile clients. Compared with the state-of-the-art framework for privacy-preserving MLaaS, SPOT achieves up to 2x memory utilization boost and a speedup up to 3x on computation time for modem neural networks such as ResNet and VGG.
Nowadays, load balancing is more and more important for huge data centers. Most data centers usually adopt a software-based load balancer (LB), which consumes too many server resources and does not scale with the incr...
详细信息
ISBN:
(纸本)9798350386066;9798350386059
Nowadays, load balancing is more and more important for huge data centers. Most data centers usually adopt a software-based load balancer (LB), which consumes too many server resources and does not scale with the increasing traffic volume. However, there is an increased demand for layer-4 load balancers, which need to check more bits in packet headers. Although hardware-based load balancers can meet the requirements, they are much more expensive. During the past years, data centers frequently update their devices, including kinds of switches. Thus, there exists a huge number of obsolete switches, some of which are in good condition and equipped with high-performance storage like SRAM (Static Random Access Memory). In this paper, we proposed BCLB (Bit-based Collaborative Load Balancer), a new cooperative LB, which builds more powerful load balancing based on existing switches. Different with previous cooperative mechanisms that distribute rules to different LBs, BCLB lets switches cooperate based on bits. Many switches along the data path check different bits, and cooperatively check all bits in 5-tuple (which would be 296 bits in IPv6) for layer-4 packet header. In this way, rule updating will not influence the load balancing in BCLB. To optimize the performance of BCLB, we formulate the problem as an optimization problem, then we propose a dynamic programming algorithm to solve it. Finally, we conduct comprehensive simulations using both real-world traffic datasets and a P4-based prototype, the results show that BCLB performs much better than previous rule-based cooperative schemes.
Motivated by a wide range of applications from parallel computing to distributed learning, we study distributed online load balancing among multiple workers. We aim to minimize the pointwise maximum over the workers&#...
详细信息
ISBN:
(纸本)9798350339864
Motivated by a wide range of applications from parallel computing to distributed learning, we study distributed online load balancing among multiple workers. We aim to minimize the pointwise maximum over the workers' local cost functions. We propose a novel algorithm termed distributed Online Load Balancing with rIsk-averse assistancE (DOLBIE), which jointly considers the worker heterogeneity and system dynamics. The workload is distributed to workers in an online manner, where the underloaded workers learn to provide an appropriate amount of assistance to the most overloaded worker for the next online round without making themselves overwhelmed. In DOLBIE, all workers participate in updating the workload simultaneously, and no computationally intensive gradient or projection calculation is required. DOLBIE can be implemented in both the master-worker and fully-distributed architectures. We analyze the worst-case performance of DOLBIE by deriving an upper bound on its dynamic regret. We further demonstrate the application of DOLBIE to online batch-size tuning in distributed machine learning. Our experimental results show that, in comparison with state-of-the-art alternatives, DOLBIE can substantially speed up the training process and reduce the workers' idle time.
The proceedings contain 67 papers. The topics discussed include: end-to-end gesture recognition framework for the identification of allergic rhinitis symptoms;semi-supervised multi-source domain adaptation in wearable...
ISBN:
(纸本)9781665495127
The proceedings contain 67 papers. The topics discussed include: end-to-end gesture recognition framework for the identification of allergic rhinitis symptoms;semi-supervised multi-source domain adaptation in wearable activity recognition;real-time human pose estimation at the edge for gait analysis at a distance;SELF-CARE: selective fusion with context-aware low-power edge computing for stress detection;publishing asynchronous event times with pufferfish privacy;tradeoff between accuracy and message complexity for approximate data aggregation;low-power distinct sum for wireless sensor networks;a software-defined underwater visible light communication testbed;a differential BCG sensor system for long term health monitoring experiment on the ISS;network economics-based crowdsourcing in UAV-assisted smart cities environments;and cost-aware inference of bovine respiratory disease in calves using precision livestock technology.
暂无评论