Date of Award
Doctor of Philosophy (PhD)
Dr. Kuang-Ching Wang, Committee Chair
Dr. Harlan Russell
Dr. Richard Brooks
Dr. Jim Martin
With the recent rise in cloud computing, applications are routinely accessing and interacting with data on remote resources. Interaction with such remote resources for the operation of media-rich applications in mobile environments is also on the rise. As a result, the performance of the underlying network infrastructure can have a significant impact on the quality of service experienced by the user. Despite receiving significant attention from both academia and industry, computer networks still face a number of challenges. Users oftentimes report and complain about poor experiences with their devices and applications, which can oftentimes be attributed to network performance when downloading or uploading application data. This dissertation investigates problems that arise with data movement across computer networks and proposes novel solutions to address these issues through software defined networking (SDN). SDN is lauded to be the paradigm of choice for next generation networks. While academia explores use cases in various contexts, industry has focused on data center and wide area networks. There is a significant range of complex and application-specific network services that can potentially benefit from SDN, but introduction and adoption of such solutions remains slow in production networks. One impeding factor is the lack of a simple yet expressive enough framework applicable to all SDN services across production network domains. Without a uniform framework, SDN developers create disjoint solutions, resulting in untenable management and maintenance overhead. The SDN-based solutions developed in this dissertation make use of a common agent-based approach. The architecture facilitates application-oriented SDN design with an abstraction composed of software agents on top of the underlying network. There are three key components modern and future networks require to deliver exceptional data transfer performance to the end user: (1) user and application mobility, (2) high throughput data transfer, and (3) efficient and scalable content distribution. Meeting these key components will not only ensure the network can provide robust and reliable end-to-end connectivity, but also that network resources will be used efficiently. First, mobility support is critical for user applications to maintain connectivity to remote, cloud-based resources. Today's network users are frequently accessing such resources while on the go, transitioning from network to network with the expectation that their applications will continue to operate seamlessly. As users perform handovers between heterogeneous networks or between networks across administrative domains, the application becomes responsible for maintaining or establishing new connections to remote resources. Although application developers often account for such handovers, the result is oftentimes visible to the user through diminished quality of service (e.g. rebuffering in video streaming applications). Many intra-domain handover solutions exist for handovers in WiFi and cellular networks, such as mobile IP, but they are architecturally complex and have not been integrated to form a scalable, inter-domain solution. A scalable framework is proposed that leverages SDN features to implement both horizontal and vertical handovers for heterogeneous wireless networks within and across administrative domains. User devices can select an appropriate network using an on-board virtual SDN implementation that manages available network interfaces. An SDN-based counterpart operates in the network core and edge to handle user migrations as they transition from one edge attachment point to another. The framework was developed and deployed as an extension to the Global Environment for Network Innovations (GENI) testbed; however, the framework can be deployed on any OpenFlow enabled network. Evaluation revealed users can maintain existing application connections without breaking the sockets and requiring the application to recover. Second, high throughput data transfer is essential for user applications to acquire large remote data sets. As data sizes become increasingly large, often combined with their locations being far from the applications, the well known impact of lower Transmission Control Protocol (TCP) throughput over large delay-bandwidth product paths becomes more significant to these applications. While myriads of solutions exist to alleviate the problem, they require specialized software and/or network stacks at both the application host and the remote data server, making it hard to scale up to a large range of applications and execution environments. This results in high throughput data transfer that is available to only a select subset of network users who have access to such specialized software. An SDN based solution called Steroid OpenFlow Service (SOS) has been proposed as a network service that transparently increases the throughput of TCP-based data transfers across large networks. SOS shifts the complexity of high performance data transfer from the end user to the network; users do not need to configure anything on the client and server machines participating in the data transfer. The SOS architecture supports seamless high performance data transfer at scale for multiple users and for high bandwidth connections. Emphasis is placed on the use of SOS as a part of a larger, richer data transfer ecosystem, complementing and compounding the efforts of existing data transfer solutions. Non-TCP-based solutions, such as Aspera, can operate seamlessly alongside an SOS deployment, while those based on TCP, such as wget, curl, and GridFTP, can leverage SOS for throughput improvement beyond what a single TCP connection can provide. Through extensive evaluation in real-world environments, the SOS architecture is proven to be flexibly deployable on a variety of network architectures, from cloud-based, to production networks, to scaled up, high performance data center environments. Evaluation showed that the SOS architecture scales linearly through the addition of SOS â€œagentsâ€ to the SOS deployment, providing data transfer performance improvement to multiple users simultaneously. An individual data transfer enhanced by SOS was shown to have increased throughput nearly forty times the same data transfer without SOS assistance. Third, efficient and scalable video content distribution is imperative as the demand for multimedia content over the Internet increases. Current state of the art solutions consist of vast content distribution networks (CDNs) where content is oftentimes hosted in duplicate at various geographically distributed locations. Although CDNs are useful for the dissemination of static content, they do not provide a clear and scalable model for the on demand production and distribution of live, streaming content. IP multicast is a popular solution to scalable video content distribution; however, it is seldom used due to deployment and operational complexity. Inspired from the distributed design of todays CDNs and the distribution trees used by IP multicast, a SDN based framework called GENI Cinema (GC) is proposed to allow for the distribution of live video content at scale. GC allows for the efficient management and distribution of live video content at scale without the added architectural complexity and inefficiencies inherent to contemporary solutions such as IP multicast. GC has been deployed as an experimental, nation-wide live video distribution service using the GENI network, broadcasting live and prerecorded video streams from conferences for remote attendees, from the classroom for distance education, and for live sporting events. GC clients can easily and efficiently switch back and forth between video streams with improved switching latency latency over cable, satellite, and other live video providers. The real world dep loyments and evaluation of the proposed solutions show how SDN can be used as a novel way to solve current data transfer problems across computer networks. In addition, this dissertation is expected to provide guidance for designing, deploying, and debugging SDN-based applications across a variety of network topologies.
Izard, Ryan, "Data Movement Challenges and Solutions with Software Defined Networking" (2017). All Dissertations. 1910.