Registration (8:00 - 11:00) (lobby of Osquldas väg 6A)

Opening and Welcome (Chair: Naeem Khademi)

(9:00 - 9:20)

Session #1 (Chair: Åke Arvidsson)

Kiran Yedugundla, Per Hurtig and Anna Brunstrom, "Probe or Wait: Handling tail losses using Multipath TCP" (30 min)

Abstract -- Packet losses are known to affect the performance of latency sensitive applications in the Internet such as media streaming and gaming. Transport protocols recover from packet loss in order to provide reliable end to end communication and improving the quality of user experience. The efficiency of loss recovery greatly influences the completion time of flows. In this paper we focus on the state of the art loss recovery mechanisms for TCP and Multipath TCP. We use controlled tail loss scenarios to evaluate the performance of loss recovery mechanisms. Based on the observations, we propose an enhancement to the tail loss recovery in Multipath TCP to improve the loss recovery time. Our experiment results show consistent end to end latency performance improvement in considered scenarios.

Tor Tangenes, David Hayes, Andreas Petlund and David Ros, "Evaluating CAIA Delay Gradient as a Candidate for Deadline-Aware Less-than-Best-Effort Transport" (30 min)

Abstract -- Less-than-best-effort (LBE) congestion control offers a low-priority service for applications tolerant to high latency and low throughput, like peer-to-peer file transfers or automatic software updates. There are, however, situations where it would be beneficial for the application to specify a soft deadline for task completion. Examples of such situations could be completion of backup tasks or synchronisation between CDN data centres. Since network conditions change over time, a deadline-aware LBE (DA-LBE) congestion control would need the ability to dynamically adapt how aggressively it competes for capacity to meet the soft deadline, trading low-priority behaviour for timeliness. One candidate that shows promise as a LBE congestion control is CAIA Delay Gradient (CDG). CDG uses changes in measured end-to-end delay to control the congestion window. CDG has several parameters that might help tune its ``aggressiveness'' in a way that might help achieve the goal of DA-LBE congestion control. We have evaluated CDG in order to establish how it can be tuned to exhibit different degrees of LBE behaviour under varying network conditions. Our results show that it is possible to control CDG to vary its aggressiveness in a consistent way, making it a prime candidate to implement a DA-LBE congestion control system.

Vladimir Marbukh, "Fragility Risks of Low Latency Dynamic Queuing in Large-Scale Clouds: Complex System Perspective" (30 min)

Abstract -- Economic, reliability, and low latency benefits of cloud computing model are the result of high level of dynamic resource sharing, made possible by high degree of system interconnectivity. This paper suggests more nuanced view of the effect of interconnectivity and resource sharing since interconnectivity may contribute to system fragility, which is associated with a possibility of abrupt/discontinuous system transition to persistent overloaded regime with unacceptably high delays. This possibility is a result of the same system interconnectivity, which enables not only desirable but also undesirable load transfer throughout the networked system. Our results indicate that benefits of dynamic queuing disciplines, including low latency, can be realized provided that the relevant systemic risks of undesirable contagion are eliminated or mitigated. Due to intractability of conventional performance models of large-scale interconnected systems, we use the "Complex Systems" methodology , e.g., mean-field approximate performance model.

Coffee Break

(10:50 - 11:15)

Workshop Panel Discussion: “Approaches to address the ossification of the Internet transport layer and challenges ahead”

Panelists:

           - Dr. Dirk Kutscher (Chief Technical Officer, Huawei) slides

           - Dr. Colin Perkins (Senior Lecturer, University of Glasgow) slides

           - Dr. Gorry Fairhurst (Professor, University of Aberdeen) (Panel Moderator) slides

(11:15 - 12:30)

Lunch Break

(12:30 - 13:45) (restaurant of Osquldas väg 6A)

Session #2 (Chair: Naeem Khademi)

Brian Trammell, Colin Perkins and Mirja Kuehlewind, "Post Sockets: Towards an Evolvable Network Transport Interface" (30 min)

Abstract -- The traditional Sockets API is showing its age, and no longer provides effective support for modern networked applications. This has led to a proliferation of non-standard extensions, alternative APIs, and workarounds that enable new features and allow applications to make good use of the network, but are difficult to use, and require expert knowledge that is not widespread. In this paper, we present Post Sockets, a proposed new standard network API, that is designed to support modern network transport protocols and features, while raising the level of abstraction and enhancing usability. Specifically, Post Sockets aims to give portable applications the ability to use a clear, messages based, interface to multi-path and multi-stream transports, rendezvous and connection racing, and fast connection re-establishment.

Tom Jones, Gorry Fairhurst and Colin Perkins, "Raising the Datagram API to Support Transport Protocol Evolution" (30 min)

Abstract -- Some application developers can wield huge resources to build new transport protocols, for these developers the UDP Socket API is perfectly fine. They have access to large test beds and sophisticated tools. Many developers do not have these resources. This paper presents a new higher Datagram API that is for everyone else, this has an advantage of offerings a clear evolutionary path to support new requirements. This new API is needed to move forward the base of the system, allowing developers with limited resources to evolve their applications while accessing new Internet functions and network services.

Felix Weinrank and Michael Tuexen, "Transparent Flow Mapping for NEAT" (30 min)

Abstract -- The NEAT library provides application developers with a unified and platform independent API for network communication, regardless of the underlying network protocol. NEAT's abstraction layer approach allows the integration of new network protocols and transport features, transparently to the user. With QUIC, RTMFP and WebRTC, several widely deployed protocols make use of mapping multiple data streams to a single transport connection. However, the usage of multiplexing requires application developers to spend additional effort and has to be supported by both endpoints. This paper describes an approach to integrate multiplexing functionality into the NEAT library, giving application developers a simple way to use the benefits of mapping multiple data streams to a single transport connection without additional coding effort. We describe our considerations about feature negotiation, connection handling and data transmission for multiplexed data streams, an introduction to the NEAT library, the implementation details as well as measurement results and future steps.

Coffee Break

(15:15 - 15:30)

Session #3 (Chair: Stein Gjessing)

Rohit P. Tahiliani and Hitesh Tewari, "Implementation of PI2 Queuing Discipline for Classic TCP Traffic in ns-3" (30 min)

Abstract -- This paper presents the implementation and validation of PI2 Active Queue Management (AQM) algorithm in ns-3. PI2 provides an alternate design and implementation to Proportional Integral controller Enhanced (PIE) algorithm without affecting the performance benefits it provides in tackling the problem of bufferbloat. Bufferbloat is a situation arising due to the presence of large unmanaged buffers in the network. It results in increased latency and therefore, degrades the performance of delay-sensitive traffic. PIE algorithm tries to minimize the queuing delay by auto-tuning its control parameters. However, with PI2, this auto-tuning is replaced by just squaring the packet drop probability. In this paper, we implement a model for PI2 in ns-3 and verify its correctness by comparing the results obtained from it to those obtained from the PIE model in ns-3. The results indicate that PI2 offers a simple design and achieves similar or at times better responsiveness and stability than PIE.

Mohammad Rajiullah, A. C. Mohideen, Felix Weinrank, Raffaello Secchi, Gorry Fairhurst and Anna Brunstrom, "Understanding Multistreaming for Web Traffic: An Experimental Study" (30 min)

Abstract -- This paper explores the design trade-offs needed for an Internet transport protocol to effectively support web access over HTTP/1.1. It explores use of a multistreaming transport protocol mechanism and studies these using a practical methodology utilising the transport features of TCP and SCTP. This is used to evaluate the relative benefit of key transport mechanisms and analyse how these impact web access performance. Our conclusions help identify the root causes of performance impairments and suggest appropriate choices when selecting a suitable transport protocol.

Concluding remarks (Chair: Åke Arvidsson)

(16:30 - 16:45)