1,721,078 research outputs found
Taming tail latency for erasure-coded, distributed storage systems
Get PDFNowadays, in distributed storage systems, long tails of responsible time are of particular concern. Modern large companies like Bing, Facebook and Amazon Web Service show that 99.9th percentile response times being orders of magnitude worse than the mean. With the advantages of maintaining high data reliability and ensur- ing enough space eciency, erasure code has become a popular storage method in distributed storage systems. However, due to the lack of mathematical models for analyzing erasure-coded based distributed storage systems, taming tail latency is still an open problem. In this research, we quantify tail latency in such systems by deriving a closed upper bounds on tail latency for general service time distribution and heterogeneous files. Later we specified service time to shifted exponentially distributed. Based on this model, we developed an optimization problem to minimize weighted tail latency probability of deriving all files. We propose an alternating minimization algorithm for this problem. Our simulation results have shown significant reduction on tail latency of erasure-coded distributed storage systems with realistic environment workload
Fast Computation of Wide Neural Networks
No full textThe recent advances in articial neural networks have demonstrated competitive performance of deep neural networks (and it is comparable with humans) on tasks like image classication, natural language processing and time series classication. These large scale networks pose an enormous computational challenge, especially in resource constrained devices. The current work proposes a targeted-rank based framework for accelerated computation of wide neural networks. It investigates the problem of rank-selection for tensor ring nets to achieve optimal network compression. When applied to a state of the art wide residual network, namely WideResnet, the framework yielded a signicant reduction in computational time. The optimally compressed non-parallel WideResnet is faster to compute on a CPU by almost 2x with only 5% degradation in accuracy when compared to a non-parallel implementation of uncompressed WideResnet.</div
Iot Based Low-Cost Precision Indoor Farming
No full textThere is a growing demand for indoor farm management systems that can track plant growth, allow automatic control and aid in real-time decision making. Internet of Thing (IoT)-based solutions are being applied to meet these needs and numerous researchers have created prototypes for meeting specific needs using sensors, algorithms, and automations. However, limited studies are available that report on comprehensive large-scale experiments to test various aspects related to availability, scalability and reliability of sensors and actuators used in low-cost indoor farms. The purpose of this study was to develop a low-cost, IoT devices driven indoor farm as a testbed for growing microgreens and other experimental crops. The testbed was designed using off-the-shelf sensors and actuators for conducting research experiments, addressing identified challenges, and utilizing remotely acquired data for developing an intelligent farm management system. The sensors were used for collecting and monitoring electrical conductivity (EC), pH and dissolved oxygen (DO) levels of the nutrient solution, light intensity, environmental variables, and imagery data. The control of light emitting diodes (LEDs), irrigation pumps, and camera modules was carried out using commercially available components. All the sensors and actuators were remotely monitored, controlled, and coordinated using a cloud-based dashboard, Raspberry Pis, and Arduino microcontrollers. To implement a reliable, real-time control of actuators, edge computing was used as it helped in minimizing latency and identifying anomalies.Decision making about overall system performance and harvesting schedule was accomplished by providing alerts on anomalies in the sensors and actuators and through installation of cameras to predict yield of microgreens, respectively. A split-plot statistical design was used to evaluate the effect of lighting, nutrition solution concentration, seed density, and day of harvest on the growth of microgreens. This study complements and expands past efforts by other researchers on building a low cost IoT-based indoor farm. While the experience with the testbed demonstrates its real-world potential of conducting experimental research, some major lessons were learnt along the way that could be used for future enhancements
Service-level based response by assignment and order processing for warehouse automation
Get PDFAlong with tremendous growth of online sales in this Internet era, unprecedented intensive competition in shortening the delivery time of orders has been occurring among several major online retailers. On the other hand, the idea of customer-oriented service creates a trend of diversified pricing strategy. Different price options are offered to cater to diversified needs of customers. It has become an urgent need for online sales industries to provide the differentiated service levels for different classes of customers with different priorities based on the charging prices and resource constraints of the supply network. In response to the challenges mentioned above, this thesis focuses on providing differentiated service levels to different customers within the warehouse automation system, which is the key point of the supply network. To concentrate on the research topic, the process of a user’s order in warehouse automation system is broken down into the waiting process and retrieving process, which is related to order processing policy and storage assignment method respectively. Priority Based Turn-over Rate (PBTR) storage assignment method, Priority Based Weighted Queuing (PBWQ) policy and joint optimization of storage assignment and PBWQ policy are proposed, developed, explored and validated in this thesis. Utility function of charging price and order processing time is developed to measure the performances of the proposed methods. Compared with the classical turn over rate assignment method, PBTR has 23.21% of improvement under the measurement of utility function, when different classes of customers have different needs for products. PBWQ improves the system performance by 18.15% compared with First-Come-First-Serve (FCFS) policy under baseline setting of experiments. Joint optimization of storage assignment and PBWQ policy has the improvement of 19.64% in system performance compared with the baseline system which applies both classical storage assignment method and FCFS order processing policy
A Quantitative Framework for CDN-Based Over-The-Top Video Streaming Systems
No full textThe demand for global video has been burgeoning across industries. With the expansion and improvement of video-streaming services, cloud-based video is evolving into a necessary feature of any successful business for reaching internal and external audiences. Over-the-top (OTT) video streaming, e.g., Netflix and YouTube, has been dominating the global IP traffic in recent years. More than 50% of OTT video traffic are now delivered through content distribution networks (CDNs).Even though multiple solutions have been proposed for improving congestion in the CDN system, managing the ever-increasing traffic requires a fundamental understanding of the system and the different design flexibilities (control knobs) to make the best use of the hardware limitations. In Addition, there is no analytical understanding for the key quality of experience (QoE) attributes (stall duration, average quality, etc.) for video streaming when transmitted using CDN-based multi-tier infrastructure, which is the focus of this thesis. The key contribution of this thesis is to provide a white-box analytical understanding of the key QoE attributes of the enduser in cloud storage systems, which can be used to systematically address the choppy user experience and have optimized system designs. The first key design involves the scheduling strategy, that chooses the subset of CDN servers to obtain the content. The second key design involves the quality of each video chunk. The third key design involves deciding which contents to cache at the edge routers and which content needs to be stored at the CDN. Towards solving these challenges, this dissertation is divided into three parts. Part 1 considers video streaming over distributed systems where the video segments are encoded using an erasure code for better reliability. Part 2 looks at the problem of optimizing the tradeoff between quality and stall of the streamed videos. In Part 3, we consider caching partial contents of the videos at the CDN as well as at the edge-routers to further optimize video streaming services.We present a model for describing a today’s representative multi-tier system architecture for video streaming applications, typically composed of a centralized origin server, several CDN sites and edge-caches. Our model comprehensively considers the following factors: limited caching spaces at the CDN sites and edge-routers, allocation of CDN for a video request, choice of different ports from the CDN, and the central storage and bandwidth allocation. With this model, we optimize different quality of experience (QoE) measures and present novel, yet efficient, algorithms to solve the formulated optimization problems. Our extensive simulation results demonstrate that the proposed algorithms significantly outperform the state-of-the-art strategies. We take one step further and implement a small-scale video streaming system in a real cloud environment, managed by Openstack, and validate our results
Flexpool: A Distributed Model-Free Deep Reinforcement Learning Algorithm for Joint Passengers & Goods Transportation
No full textThe growth in online goods delivery is causing a dramatic surge in urban vehicle traffic from last-mile deliveries. On the other hand, ride-sharing has been on the rise with the success of ride-sharing platforms and increased research on using autonomous vehicle technologies for routing and matching. The future of urban mobility for passengers and goods relies on leveraging new methods that minimize operational costs and environmental footprints of transportation systems.This paper considers combining passenger transportation with goods delivery to improve vehicle-based transportation. Even though the problem has been studied with model-based approaches where the dynamic model of the transportation system environment is defined, model-free approaches where the dynamics of the environment are learned by interaction have been demonstrated to be adaptable to new or erratic environment dynamics. FlexPool is a distributed model-free deep reinforcement learning algorithm that jointly serves passengers & goods workloads by learning optimal dispatch policies from its interaction with the environment. The model-free algorithm (as opposed to a model-based one) is an algorithm which does not use the transition probability distribution (and the reward function) associated with the Markov decision process (MDP). The proposed algorithm pools passengers for a ride-sharing service and delivers goods using a multi-hop routing method. These flexibilities decrease the fleet’s operational cost and environmental footprint while maintaining service levels for passengers and goods. The dispatching algorithm based on deep reinforcement learning is integrated with an efficient matching algorithm for passengers and goods. Through simulations on a realistic urban mobility platform, we demonstrate that FlexPool outperforms other model-free settings in serving the demands from passengers & goods. FlexPool achieves 30% higher fleet utilization and 35% higher fuel efficiency in comparison to (i) model-free approaches where vehicles transport a combination of passengers & goods without the use of multi-hop transit, and (ii) model-free approaches where vehicles exclusively transport either passengers or goods
Multi-Dimensional Data Analytics and Deep Learning Via Tensor Networks
Get PDFWith the booming of big data and multi-sensor technology, multi-dimensional data, known as tensors, has demonstrated promising capability in capturing multi- dimensional correlation via efficiently extracting the latent structures, and drawn considerable attention in multiple disciplines such as image processing, recommender system, data analytics, etc. In addition to the multi-dimensional nature of real data, artificially designed tensors, referred as layers in deep neural networks, have also been intensively investigated and achieved the state-of-the-art performance in imaging processing, speech processing, and natural language understanding. However, algorithms related with multi-dimensional data are unfortunately expensive in computation and storage, thus limiting its application when the computational resources are limited. Although tensor factorization has been proposed to reduce the dimensionality and alleviate the computational cost, the trade-off among computation, storage, and performance has not been well studied. To this end, we first investigate an efficient dimensionality reduction method using a novel Tensor Train (TT) factorization. In particular, we propose a Tensor Train Principal Component Analysis (TT-PCA) and a Tensor Train Neighborhood Preserving Embedding (TT-NPE) to project data onto a Tensor Train Subspace (TTS) and effectively extract the discriminative features from the data. Mathematical analysis and simulation demonstrate TT-PCA and TT-NPE achieve better trade-off among computation, storage, and performance than the bench-mark tensor-based dimensionality reduction approaches. We then extend the TT factorization into general Tensor Ring (TR) factorization and propose a tensor ring completion algorithm, which can utilize 10% randomly observed pixels to recover the gunshot video at an error rate of only 6.25%. Inspired by the novel trade-off between model complexity and data representation, we introduce a Tensor Ring Nets (TRN) to compress the deep neural networks significantly. Using the benchmark 28-layer WideResNet architectures, TRN is able to compress the neural network by 243× with only 2.3% degradation in Cifar10 image classification
A Distributed Model-Free Algorithm for Multi-Hop Ride-Sharing Using Deep Reinforcement Learning
No full textThe growth of autonomous vehicles, ridesharing systems, and self driving technology will bring a shift in the way ride hailing platforms plan out their services. However, these advances in technology coupled with road congestion, environmental concerns, fuel usage, vehicles emissions, and the high cost of the vehicle usage have brought more attention to better utilize the use of vehicles and their capacities. In this paper, we propose a novel distributed multi-hop ride-sharing (MHRS) algorithm that uses deep reinforcement learning to learn optimal vehicle dispatch and matching decisions by interacting with the external environment. By allowing customers to transfer between vehicles, i.e., ride with one vehicle for sometime and then transfer to another one, MHRS helps in attaining 30% lower cost and 20% more efficient utilization of fleets, as compared to the ride-sharing algorithms. This flexibility of multi-hop feature gives a seamless experience to customers and ride-sharing companies, and thus improves ride-sharing services
Deep Learning Based Models for Novelty Adaptation in Autonomous Multi-Agent Systems
No full textAutonomous systems are often deployed in dynamic environments and are challenged with unexpected changes (novelties) in the environments where they receive novel data that was not seen during training. Given the uncertainty, they should be able to operate without (or with limited) human intervention and they are expected to (1) Adapt to such changes while still being effective and efficient in performing their multiple tasks. The system should be able to provide continuous availability of its critical functionalities. (2) Make informed decisions independently from any central authority. (3) Be Cognitive: learns the new context, its possible actions, and be rich in knowledge discovery through mining and pattern recognition. (4) Be Reflexive: reacts to novel unknown data as well as to security threats without terminating on-going critical missions. These characteristics combine to create the workflow of autonomous decision-making process in multi-agent environments (i.e.,) any action taken by the system must go through these characteristic models to autonomously make an ideal decision based on the situation.In this dissertation, we propose novel learning-based models to enhance the decision-making process in autonomous multi-agent systems where agents are able to detect novelties (i.e., unexpected changes in the environment), and adapt to it in a timely manner. For this purpose, we explore two complex and highly dynamic domains (1) Transportation Networks (e.g., Ridesharing application): where we develop AdaPool: a novel distributed diurnal-adaptive decision-making framework for multi-agent autonomous vehicles using model-free deep reinforcement learning and change point detection. (2) Multi-agent games (e.g., Monopoly): for which we propose a hybrid approach that combines deep reinforcement learning (for frequent but complex decisions) with a fixed-policy approach (for infrequent but straightforward decisions) to facilitate decision-making and it is also adaptive to novelties. (3) Further, we present a domain agnostic approach for decision making without prior knowledge in dynamic environments using Bootstrapped DQN. Finally, to enhance security of autonomous multi-agent systems, (4) we develop a machine learning based resilience testing of address randomization moving target defense. Additionally, to further improve the decision-making process, we present (5) a novel framework for multi-agent deep covering option discovery that is designed to accelerate exploration (which is the first step of decision-making for autonomous agents), by identifying potential collaborative agents and encouraging visiting the under-represented states in their joint observation space
A Framework for Training Spiking Neural Networks Using Evolutionary Algorithms and Deep Reinforcement Learning
No full textIn this work two novel frameworks, one using evolutionary algorithms and another using Reinforcement Learning for training Spiking Neural Networks are proposed and analyzed. A novel multi-agent evolutionary robotics (ER) based framework, inspired by competitive evolutionary environments in nature, is demonstrated for training Spiking Neural Networks (SNN). The weights of a population of SNNs along with morphological parameters of bots they control in the ER environment are treated as phenotypes. Rules of the framework select certain bots and their SNNs for reproduction and others for elimination based on their efficacy in capturing food in a competitive environment. While the bots and their SNNs are given no explicit reward to survive or reproduce via any loss function, these drives emerge implicitly as they evolve to hunt food and survive within these rules. Their efficiency in capturing food as a function of generations exhibit the evolutionary signature of punctuated equilibria. Two evolutionary inheritance algorithms on the phenotypes, Mutation and Crossover with Mutation along with their variants, are demonstrated. Performances of these algorithms are compared using ensembles of 100 experiments for each algorithm. We find that one of the Crossover with Mutation variants promotes 40% faster learning in the SNN than mere Mutation with a statistically significant margin. Along with an evolutionary approach to training SNNs, we also describe a novel Reinforcement Learning(RL) based framework using the Proximal Policy Optimization to train a SNN for an image classification task. The experiments and results of the framework are then discussed highlighting future direction of the work
- …
