Education & Academic Experience

Dissertation Abstract

Today, cloud services offer myriads of applications, tailor made for different users in the field of weather, health, finance, entertainment, etc. These services fulfill varying genres of user demands over the Internet. For example, these services can be live (live weather radar, ESPN Live) or on-demand services (weather forecasting, Netflix). While these applications cater to different customer requirements, it is necessary for these services to be efficient with respect to latency, scalability, robustness and quality of experience. These systems need to constantly evolve to provide the best user experience and meet the most current demands of the customer. For instance, weather forecasting systems need to make accurate predictions close to real-time in order to alert customers of an imminent disaster. Similarly, video-streaming systems consistently strive towards providing best viewing experience to the customer. As weather predictions become more fine-grained with high-resolution observations and video streaming platforms evolve with higher resolution content, the underlying system needs to improve to deliver optimum performance. Even though weather sensing and forecasting has become increasingly accurate in the last decade thanks to high-resolution radars, efficient computational algorithms, and high-performance computing facilities, there are existing challenges that could result in sub-optimal weather prediction. Flood-forecasting models are dependent on high-performance computing clusters for processing high-resolution rainfall data and their unavailability can become a bottleneck. Also, weather radars are critical infrastructure, but are often located in remote areas with poor network connectivity. Data retrieved from these radars are often delayed or lost, or even lack proper synchronization, resulting in less than ideal weather predictions. Additionally, as more and more users have access to the cloud-based video-streaming platforms, their viewing experience is often adversely affected by network conditions (congestion, limited bandwidth) and limitations of address-based Internet architectures. In this dissertation, we try to address these existing challenges by designing and implementing systems on the cloud that improve end-user experience. We optimize two cloud services: In the field of safety we improve upon weather forecasting systems and in the field of entertainment, quality of experience in video-streaming systems. We utilize, investigate and incorporate benefits of different system technologies such as parallelization, virtualization, software defined networking and information centric networking in the existing system workflow with the goal of optimizing the quality of experience. Firstly, we relieve the dependency on high-performance computing clusters for a flood-forecasting model by using cloud resources as an alternative. We then optimize the performance of this model on the cloud by parallely distributing its existing workflow, which leads to 34% reduction in rainfall processing time. The observed improvement in runtime is not only consistent with varying rainfall conditions but could also lead to faster forecasts on the cloud and potentially be extended to different weather forecast models. Secondly, we improve data-retrieval and synchronization challenges in weather sensing radars by applying Named Data Networking (NDN) for efficient content addressing and retrieval. We identify weather data based on a hierarchical naming scheme to explicitly access desired data and demonstrate that compared to a time-based windowing mechanism for radar data retrieval on top of TCP/IP, an NDN based mechanism improves data quality, reduces uncertainty, and enhances weather prediction. Lastly, we improve quality of experience in live video streaming by designing a seamless and efficient distribution system that is flexible enough to choose from multiple Internet architectures best suited for live video streaming. We highlight the benefits of such a hybrid system for live video streaming as well as present a detailed analysis with the goal to provide a high quality of experience (QoE) for the viewer. For our hybrid architecture, video streaming is supported simultaneously over TCP/IP and Named Data Networking (NDN)-based architecture via operating system and networking virtualization techniques to design a flexible system that utilizes the benefits of these varying Internet architectures. Based on a prototype we have designed and implemented maintaining efficient use of network resources, we demonstrate that in the case of live streaming, NDN achieves better QoE per client than IP and can also utilize higher than the available bottleneck bandwidth through in-network caching. Even without caching, as opposed to IP-only, our hybrid setup achieves better average bitrate and better perceived visual quality over live video streaming services. Furthermore, we present detailed analysis on ways adaptive video streaming with NDN can be further improved with respect to QoE. In this context, we identify the phenomenon of oscillation effects that adversely affects QoE in adaptive video streaming and propose solutions to reduce them.

Systems & Networks

• E&C-ENG 671: Computer Networks

Machine Learning & AI

• CS689: Machine Learning

Other

• CS590CC: Cloud Computing

Systems & Networks

• CS677: Operating Systems

Machine Learning & AI

• CS585: Intro Natural Language Processing

Theory & Other

• CS611: Advanced Algorithms
• CS646: Information Retrieval
• CS621: Adv S/W Eng: Analysis and Eval
• CS701: Advanced Topics Computer Science

Advisor: Prashant Shenoy

📅 Feb - Aug 2016 | 📍 UMass Amherst

Real-time Water-quality monitoring: Built a cost-effective & real-time solution for alerting users with inconsumable water quality.

Advisor: J. Elliot B. Moss

📅 Jan - Aug 2015 | 📍 UMass Amherst

Java Virtual Machine Bytecode Optimization: The work involved identifying and removing redundancies from compiled JAVA class files that effectively reduced the runtime of JVM.