Performance, Availability, and Scalability in Open Networking Platforms for Internet Services
Time: Fri 2021-09-03 13.00
Location: zoom link for online defense (English)
Subject area: Information and Communication Technology
Doctoral student: Voravit Tanyingyong , Kommunikationssystem, CoS
Opponent: Professor Raimo Kantola, Aalto University
Supervisor: Associate Professor Peter Sjödin, Kommunikationssystem, CoS; Markus Hidell, Kommunikationssystem, CoS
Abstract
Advancement in technologies brings about new demands and services. In parallel, modern commodity off-the-shelf (COTS) hardware has become very capable yet affordable. It is often used in combination with Free/Libre and Open Source Software (FLOSS) to build open networking platforms, which offer alternatives to commercial network devices.
In this thesis, we study open networking platforms for Internet services in different use cases. We focus on different aspects of the quality of Internet services in three classes of network devices, including performance aspects in high-end network devices for access networks in the context of software-defined networking (SDN), availability aspects in mid-range network devices for home networks in the context of healthcare services, and scalability aspects in low-end network devices for the Internet of Things (IoT).
For the performance aspect in an SDN context, we design, implement, and evaluate open networking platforms with an OpenFlow forwarding engine. Our architecture utilizes the hardware classification on a network interface card (NIC) to improve OpenFlow lookup performance. Moreover, we devise a strategy to utilize multi-core architecture to extend our open networking platform to perform other tasks beyond solely routing.
For the availability aspect in a healthcare context, we combine mission critical healthcare services with a traditional residential gateway. Our mechanism to enhance availability utilizes the bidirectional forwarding detection (BFD) protocol to enable redundancy with fast failover and dedicated system resources to provide performance isolation for BFD control messages.
We consider the scalability aspect for IoT in two scenarios, namely IoT actuation in a smart grid context and IoT sensing in a smart city context. For IoT actuation, we introduce IoT-grid, a programmable small-scale, DC power grid that utilizes embedded IoT technology to enable remote control and monitoring functions. We observe that IoT-enabled grid components have long processing delays and, thus, propose a mechanism to improve the scalability based on sending a burst of requests with scheduled responses.
For IoT sensing, we design and develop an IoT-based air quality monitoring testbed that offers reliable data delivery. Moreover, we propose dynamic sinks, sensor nodes that can serve as on-demand sinks, to improve the scalability of IoT sensing systems while maintaining low energy consumption.
These use cases demonstrate that open networking platforms can meet service quality demands for various applications and offer viable alternatives to commercial network devices.