Internet Engineering at the University of Aberdeen

Research in Internet Engineering at the University of Aberdeen led by Prof. Gorry Fairhurst and Dr Raffaello Secchi and the wider team, has focussed on addressing key Internet Engineering challenges and has played a fundamental role in enhancing the efficiency of Internet systems through collaborations in a range of projects (see below).

The day-to-day operation of the Internet depends on voluntary adherence of vendors and operators to standard protocols and procedures specified by the Internet Engineering Task Force (IETF). This work spans the design of Internet equipment, measurement of Internet infrastructure, to the design of the protocols that control data transport across Internet paths. RFC specifications are crucial to the day-to-day running and expansion of the Internet. Although the commercial gain yielded by such open standards is not measurable, these standards are widely used at all stages in Internet service delivery: by networking equipment designers, by network and service operators, in data centres and as a basis for enterprise. RFCs are also used as the basis for government procurement of equipment and services, and underpin other standardisation, including cellular mobile standards for 5G technology.

Key research themes in Internet Engineering

Evolving the Internet Transport System and defining a new Transport Interface

Our research is exploring the networking architecture for end to end communications.

Working with key industry partners including Mozilla, CISCO and Celerway, underpinned the design of a new architecture for the transport system (TAPS). This is addressing key obstacles faced by Internet innovators by 1) lowering the barrier to application development through a new open transport system, that allows developers to automatically select a suitable protocol for an Internet application and to specify the options required; and 2) providing an architectural change where new transport services can seamlessly be integrated.

The TAPS archiecture, decouples the transport service from the design of the underlying transport protocol. This eliminates previous obstacles to deploying new methods, enabling evolution within the transport system, and allowing developers to benefit from new methods wherever they become available. Partners in the NEAT Project (see below) proposed a new system architecture and later contributed this to the IETF. An Open Source research implementation demonstrated the feasibility of this approach. Fairhurst and his team implemented the Datagram subsystem, and then defined the interaction with the network layer, resulting in further standards contributions around differentiated services, path MTU discovery, management of encrypted networks, etc.

Our research team has designed innovative research tools enabling large-scale Internet measurement of the transparency of paths to various Internet protocol mechanisms. These tools are used to examine mobile and wired Internet paths (e.g. MAMI, PREC, Tor). Employing these tools measurements provide engineering reality checks to help inform evolution of the Internet Architecture and protocols.

Interent Transport over Broadband Satellite

A wide variety of enterprise and broadband satellite systems have been studied as a part of the research at the University of Aberdeen. This includes design of methods for IETF ULE, DVB GSE, DVB-RCS, and other subsystems. We work on transport protocols and their interaction with the end to end paths provided by geostationary and non geostationary satellite systems. As part of a project funded by the European Space Agency (ESA), the team analyses TCP, HTTP2, QUIC, an encrypted UDP transport, and evaluated performance of web protocols. This research is proposing changes to the IETF QUIC transport, and contributions to ensure acceptable performance of the emerging IETF QUIC protocol over broadband satellite systems.

Reducing End-to-End Internet Latency across the Internet

Research in the interaction of transport and network protocols performed an in-depth analysis of the end-to-end delay experienced by Internet users. This work in RITE and MAMI brought Internet latency to the fore as an important performance metric. Latency measures the time needed to transport packets across the network, which ultimately impacts download speed and responsiveness of communications. Evaluation of the performance of Active Queue Management (AQM) methods showed these can significantly reduce latency, without requiring investment in higher speed communications links. Explicit Congestion Notification (ECN) methods were defined that enable a router to better control the latency it introduces, resulting in the design of a new congestion control algorithm, TCP Alternative Backoff with ECN (ABE), and subsequent development of ideas by others in L4S.

Determining Appropriate Sending Behaviour for Internet Datagrams

From 2015-2018, and building on research around the User Datagram Protocol (UDP), we explored the impact of the industry move towards UDP transport by focusing on how network operators balance operational needs and user privacy concerns, to understand the implications of encryption on manageability of network traffic. Dr Fairhurst and Dr Verdicchio have worked on transport for video. Dr Secchi has studied ML mechanisms to identify QoS features from encrypted streams.

Following analysis of path failure due to packet size, our team worked with the University of Muenster to co-design a new technique, Datagram Packetization Layer PMTU Discovery, DPLPMTUD. This can automatically determine when it is safe to increase the size of packet being used, which allows a sender to detect a black hole and reduce the packet size (when current methods could fail by using full-sized packets). DPLPMTUD successfully avoids packet size black holes for 99.72% of tested IPv4 Internet paths. Using larger packets can reduce the packet rate by 17%, for 96% of paths, compared to using a default size of 1280 B. We have specified additions to IPv6 and UDP Options that assist PLMTU discovery.

List of Internet Engineering Projects