Browsing by Advisor "Zink, Michael"

Now showing 1 - 8 of 8
  • Thumbnail Image
    Publication
    AN EVALUATION OF SDN AND NFV SUPPORT FOR PARALLEL, ALTERNATIVE PROTOCOL STACK OPERATIONS IN FUTURE INTERNETS
    (2018-05) Suresh, Bhushan
    Virtualization on top of high-performance servers has enabled the virtualization of network functions like caching, deep packet inspection, etc. Such Network Function Virtualization (NFV) is used to dynamically adapt to changes in network traffic and application popularity. We demonstrate how the combination of Software Defined Networking (SDN) and NFV can support the parallel operation of different Internet architectures on top of the same physical hardware. We introduce our architecture for this approach in an actual test setup, using CloudLab resources. We start of our evaluation in a small setup where we evaluate the feasibility of the SDN and NFV architecture and incrementally increase the complexity of the setup to run a live video streaming application. We use two vastly different protocol stacks, namely TCP/IP and NDN to demonstrate the capability of our approach. The evaluation of our approach shows that it introduces a new level of flexibility when it comes to operation of different Internet architectures on top of the same physical network and with this flexibility provides the ability to switch between the two protocol stacks depending on the application.
  • Thumbnail Image
    Publication
    Efficient Scaling of a Web Proxy Cluster
    (2017-09) Zhang, Hao
    With the continuing growth in network traffic and increasing diversity in web content, web caching, together with various network functions (NFs), has been introduced to enhance security, optimize network performance, and save expenses. In a large enterprise network with more than tens of thousands of users, a single proxy server is not enough to handle a large number of requests and turns to group processing. When multiple web cache proxies are working as a cluster, they talk with each other and share cached objects by using internet cache protocol (ICP). This leads to poor scalability. This thesis describes the development of a framework that provides the efficient management of a distributed web cache. A controller is introduced into the cluster of proxy servers and becomes responsible for managing objects shared within the cluster. By obtaining a knowledge of global states from the controller, proxy servers that are working in the group do not need to query its neighbors' storage. This reduces traffic in the cluster and saves the computing resources of associated proxy servers. The evaluation on a caching proxy benchmark has shown that our approach demonstrates a superior scalability in comparison to an ICP web caching cluster.
  • Thumbnail Image
    Publication
    Evaluating Adaptive Filter Techniques for Wind Noise Removal from Infrasound Barometers
    (2024-09) Saplakoglu, Hakan
    Infrasound signals, meaning those signals below 20 Hz, are commonly found throughout nature and man-made infrastructure. These signals can be sourced and analyzed using infrasound barometers or microphones. Often, infrasound signals carry very weak power due to their source being large distances away. Infrasound sensors can be very precise and are able to record these signals, however, in the presence of any external noise, these signals are easily lost. The most prominent source of noise in an outdoor sensing environment is wind, a non-stationary noise source. In this thesis, an ultrasonic anemometer is used to measure instantaneous wind speed at the same location to where an infrasound barometer takes measurements. This thesis will show that there is a weak correlation between the anemometer noise signal and the barometer signal. Normalized least-mean-square (NLMS) adaptive filtering will be used, specifically its residual error, as the noise-cancelled output from the barometer data. It will be shown that this method is successful at reducing a portion of the wind noise in the barometer signal, however, also that this method is not entirely successful in that it is not able to uncover lost infrasound signals due to noise. To evaluate this, a separate barometer with a passive wind filter will be deployed to serve as a comparison to an ideal condition. This thesis will also show that this is a challenging problem and outline the various solutions used to overcome challenges throughout the process.
  • Thumbnail Image
    Publication
    Explorations into Machine Learning Techniques for Precipitation Nowcasting
    (2017-02) Nagarajan, Aditya
    Recent advances in cloud-based big-data technologies now makes data driven solutions feasible for increasing numbers of scientific computing applications. One such data driven solution approach is machine learning where patterns in large data sets are brought to the surface by finding complex mathematical relationships within the data. Nowcasting or short-term prediction of rainfall in a given region is an important problem in meteorology. In this thesis we explore the nowcasting problem through a data driven approach by formulating it as a machine learning problem. State-of-the-art nowcasting systems today are based on numerical models which describe the physical processes leading to precipitation or on weather radar extrapolation techniques that predict future radar precipitation maps by advecting from a sequence of past maps. These techniques, while they can perform well over very short prediction horizons (minutes) or very long horizons (hours to days), tend not to perform well over medium horizons (1-2 hours) due to lack of input data at the necessary spatial and temporal scales for the numerical prediction methods or due to the inability of radar extrapolation methods to predict storm growth and decay. Given that water must first concentrate in the atmosphere as water vapor before it can fall to the ground as rain, one goal of this thesis is to understand if water vapor information can improve radar extrapolation techniques by giving the information needed to infer growth and decay. To do so, we use the GPS-Meteorology technique to measure the water vapor in the atmosphere and weather radar reflectivity to measure rainfall. By training a machine learning nowcasting algorithm using both variables and comparing its performance against a nowcasting algorithm trained on reflectivity alone, we draw conclusions as to the predictive power of adding water vapor information. Another goal of this thesis is to compare different machine learning techniques, viz., the random forest ensemble learning technique, which has shown success on a number of other weather prediction problems, and the current state-of-the-art machine learning technique for images and image sequences, convolutional neural network (CNN). We compare these in terms of problem representation, training complexity, and nowcasting performance. A final goal is to compare the nowcasting performance of our machine learning techniques against published results for current state-of-the-art model based nowcasting techniques.
  • Thumbnail Image
    Publication
    Improving Resilience of Communication in Information Dissemination for Time-Critical Applications
    (2019-05) Deshmukh, Rajvardhan Somraj
    Severe weather impacts life and in this dire condition, people rely on communication, to organize relief and stay in touch with their loved ones. In such situations, cellular network infrastructure\footnote{We refer to cellular network infrastructure as infrastructure for the entirety of this document} might be affected due to power outage, link failures, etc. This urges us to look at Ad-hoc mode of communication, to offload major traffic partially or fully from the infrastructure, depending on the status of it. We look into threefold approach, ranging from the case where the infrastructure is completely unavailable, to where it has been replaced by make shift low capacity mobile cellular base station. First, we look into communication without infrastructure and timely, dissemination of weather alerts specific to geographical areas. We look into the specific case of floods as they affect significant number of people. Due to the nature of the problem we can utilize the properties of Information Centric Networking (ICN) in this context, namely: i) Flexibility and high failure resistance: Any node in the network that has the information can satisfy the query ii) Robust: Only sensor and car need to communicate iii) Fine grained geo-location specific information dissemination. We analyze how message forwarding using ICN on top of Ad hoc network, approach compares to the one based on infrastructure, that is less resilient in the case of disaster. In addition, we compare the performance of different message forwarding strategies in VANETs (Vehicular Adhoc Networks) using ICN. Our results show that ICN strategy outperforms the infrastructure-based approach as it is 100 times faster for 63\% of total messages delivered. Then we look into the case where we have the cellular network infrastructure, but it is being pressured due to rapid increase in volume of network traffic (as seen during a major event) or it has been replaced by low capacity mobile tower. In this case we look at offloading as much traffic as possible from the infrastructure to device-to-device communication. However, the host-oriented model of the TCP/IP-based Internet poses challenges to this communication pattern. A scheme that uses an ICN model to fetch content from nearby peers, increases the resiliency of the network in cases of outages and disasters. We collected content popularity statistics from social media to create a content request pattern and evaluate our approach through the simulation of realistic urban scenarios. Additionally, we analyze the scenario of large crowds in sports venues. Our simulation results show that we can offload traffic from the backhaul network by up to 51.7\%, suggesting an advantageous path to support the surge in traffic while keeping complexity and cost for the network operator at manageable levels. Finally, we look at adaptive bit-rate streaming (ABR) streaming, which has contributed significantly to the reduction of video playout stalling, mainly in highly variable bandwidth conditions. ABR clients continue to suffer from the variation of bit rate qualities over the duration of a streaming session. Similar to stalling, these variations in bit rate quality have a negative impact on the users’ Quality of Experience (QoE). We use a trace from a large-scale CDN to show that such quality changes occur in a significant amount of streaming sessions and investigate an ABR video segment retransmission approach to reduce the number of such quality changes. As the new HTTP/2 standard is becoming increasingly popular, we also see an increase in the usage of HTTP/2 as an alternative protocol for the transmission of web traffic including video streaming. Using various network conditions, we conduct a systematic comparison of existing transport layer approaches for HTTP/2 that is best suited for ABR segment retransmissions. Since it is well known that both protocols provide a series of improvements over HTTP/1.1, we perform experiments both in controlled environments and over transcontinental links in the Internet and find that these benefits also “trickle up” into the application layer when it comes to ABR video streaming where HTTP/2 retransmissions can significantly improve the average quality bitrate while simultaneously minimizing bit rate variations over the duration of a streaming session. Taking inspiration from the first two approaches, we take into account the resiliency of a multi-path approach and further look at a multi-path and multi-stream approach to ABR streaming and demonstrate that losses on one path have very little impact on the other from the same multi-path connection and this increases throughput and resiliency of communication.
  • Thumbnail Image
    Publication
    Lecture Video Transformation through An Intelligent Analysis and Post-processing System
    (2021-05) Wang, Xi
    Lecture videos are good sources for people to learn new things. Students commonly use online videos to explore various domains. However, some recorded videos are posted on online platforms without being post-processed due to technology and resource limitations. In this work, we focus on the research of developing an intelligent system to automatically extract essential information, including the main instructor and screen, in a lecture video in several scenarios by using modern deep learning techniques. This thesis aims to combine the extracted essential information to render the videos and generate a new layout with a smaller file size than the original one. Another benefit of using this approach is that the users may save video post-processing time and costs. State-of-the-art object detection models, an algorithm to correct screen display, tracking the instructor, and other deep learning techniques were adopted in the system to detect both the main instructor and the screen in given videos without much of the computational burden. There are four main contributions: 1. We built an intelligent video analysis and post-processing system to extract and reframe detected objects from lecture videos. 2. We proposed a post-processing algorithm to localize the frontal human torso position in processing a sequence of frames in the videos. 3. We proposed a novel deep learning approach to distinguish the main instructor from other instructors or audiences in several complex situations. 4. We proposed an algorithm to extract the four edge points of a screen at the pixel level and correct the screen display in various scenarios.
  • Thumbnail Image
    Publication
    Synchronized Object Sharings for Augmented Reality Virtual Conferencing
    (2024-05) Murray, John O.
    In the aftermath of recent global events and the gradual subsiding of the pandemic, governments and organizations worldwide are proactively preparing for future challenges. The importance of communication and connection has been underscored in recent years, especially through applications like Skype, Discord, and Zoom, as they play a pivotal role in collaboration and innovation across the world. Virtual platforms have become integral for societies to connect and collaborate, emphasizing the need for the evolution of communication methods to ensure a resilient global future. This thesis examines existing conferencing applications and provides an implementation for a basic Extended Reality (XR) conferencing application. This required implementing an XR-capable application for the HoloLens 2 as well as creating a server application that acts as a hub for all connected users, with a focus on simplicity and modality to allow for future modification and experimentation. Once these applications were created, further tests were performed to evaluate network latency, QR-code scanning metrics, and user quality of experience through a short user study. The subsequent discussion will expound on the implementation process, technical intricacies, and potential issues regarding the integration of XR conferencing systems. As we anticipate the challenges that lie ahead, the development and deployment of a basic conferencing application for the HoloLens 2 signifies a strategic step in building a more connected and resilient future. This thesis will conclude by providing insights into the attained results, offering a comprehensive understanding of the tangible outcomes and implications of this work within the context of advancing communication technologies for a post-pandemic world.
  • Thumbnail Image
    Publication
    VIRTUALIZATION OF CLOSED-LOOP SENSOR NETWORKS
    (2017-05) Kedalagudde, Priyanka Dattatri
    The existing closed-loop sensor networks are based on architectures that are designed and implemented for one specific application and require dedicated sensing and computational resources. This prevents the sharing of these networks. In this work, we propose an architecture of virtualization to allow sharing of closed-loop sensor networks. We also propose a scheduling approach that will manage requests from competing applications and evaluate their impact on system utilization against utilization achieved by more traditional, dedicated sensor networks. These algorithms are evaluated through trace-driven simulations, where the trace data is taken from CASA’s closed-loop weather radar sensor network. Results from this evaluation show that the proposed scheduling algorithms applied in a shared network result in cost savings, that are the result of being able to multiplex applications onto a single network as opposed to running each application on an dedicated sensor network.