FIT 2020 Program

Monday - June 22, 2020

09:00 - 09:15Opening and Welcome

Session Chairs: Naeem Khademi and Karl-Johan Grinnemo

09:15 - 10:45Session #1: Quality of Service

Session Chair: Naeem Khademi

  • A DRAM-friendly priority queue Internet packet scheduler implementation and its effects on TCP - Katsushi Kobayashi (30 min)

To meet various latency requirements for network applications, latency aware packet schedulers that respect per packet deadlines requested by end systems, such as least slack-time first (LSTF), have emerged. Even though latency aware schedulers are beneficial, few issues remain. First, it is not compatible with best-effort networks, such as the Internet, because of infinite buffer delays by keeping every packet without discarding even in a congestion. Second, its implementation is challenging with increasing network bandwidths due to memory device restrictions, i.e. DRAM access latencies. For the congestion issue, we presented earliest deadline first with reneging (EDFR), which can be applied to besteffort services by taking advantage of its packet drop feature. This paper discusses EDFR scheduler implementations on FPGA and its impact on TCP stack with real systems. We designed the skip-FIFO packet scheduler, which is expected to be compatible with DRAM. We implemented skip-FIFO based EDFR onto FPGA, which performed to take advantage of memory bandwidth, such as 75% of the theoretical limit of High Bandwidth Memory (HBM) DRAM. The TCP behaviors on EDFR were almost unchanged in terms of throughputs and losses even for coexisting flows requesting different latencies.

  • SRPT-ECF: challenging Round-Robin for streamaware multipath scheduling - Baptiste Jonglez, Martin Heusse and Bruno Gaujal (30 min)

Multipath TCP has long been the standard multipath transport protocol. However, the recent introduction of Multipath QUIC has changed the landscape by allowing multiple streams to coexist, bringing opportunities for further optimisation but also a new set of challenges. New stream-aware scheduling algorithms are necessary to account for this new variable. We show that, perhaps counterintuitively, serving streams using a Round-Robin strategy yields poor performance when looking at stream completion time. We then describe SRPT-ECF, our novel stream-aware multipath scheduling algorithm. We show that our algorithm is optimal in a simple network model and that it exhibits good properties on HTTP/2 traces. We then sketch how it could be implemented within Multipath QUIC to schedule web resources with HTTP/2, paving the way for low-latency multipath HTTP/3 implementations.

  • Transport Layer Efficiency and Security in IoT: RINA’s Approach - Peyman Teymoori and Toktam Ramezanifarkhani (30 min)

Technological advancements have enabled us to connect tiny things to the Internet to benefit from an automatized, easier life style using the Internet of Things (IoT). This, however, raised serious challenges, especially at the communication/networking layer of IoT. In particular, meeting both of the transport layer efficiency and security has been a serious research question and open problem. In this paper, we take a deeper look at this problem by summarizing current network stacks of IoT, and then, through some use cases, we demonstrate how the idea of recursion in networking can solve this problem; we employ the Recursive InterNetwork Architecture (RINA) as our architectural approach, and show how this idea can conquer the above challenge without sacrificing security or efficiency, and how secure multicast, a crucial requirement of IoT, can be inherently performed in IoT.

10:45 - 11:00  Coffee Break

11:00 - 11:40Keynote: HTTP/3 is coming - Daniel Stenberg

Biography: Daniel Stenberg is the founder and lead developer of the curl project. He has worked on HTTP implementations for over twenty years. He has been involved in the HTTPbis working group in IETF a long time and he worked on the HTTP stack in Firefox for several years at Mozilla. He participates in the QUIC working group and is the author of the widely read documents "HTTP2 explained" and "HTTP/3 explained". Employed by wolfSSL.

Abstract: HTTP/3 is the designated name for the coming next version of the protocol that is currently under development within the QUIC working group in the IETF. It is designed to improve in areas where HTTP/2 still has some shortcomings, primarily by changing the transport layer, and is the first major protocol to step away from TCP and instead it uses QUIC. Why is this new protocol set deemed necessary? How do they work? They change how things are sent over the network and with this comes new deployment challenges. You can join in the fun and run it and try it out right now!

11:40 - 12:40Session #2: Transport Layer

Session Chair: Karl-Johan Grinnemo

  • Rethinking ACKs at the Transport Layer - Ana Custura, Tom Jones and Gorry Fairhurst (30 min)

Acknowledgements are a core component of transport protocols and help provide reliability and congestion control.They have been used to send control information and to receive feedback about progress, to measure responsiveness and capacity of the path, and numerous other purposes. The TCP has inspired the design of many other protocols, including DCCP, SCTP and QUIC. This paper takes a long look at the way acknowledgements are used in transport protocols and what is actually needed in new protocol designs.

  • A QUIC Simulation Model for INET and its Application to the Acknowledgment Ratio Issue - Timo Völker, Ekaterina Volodina, Michael Tüxen and Erwin P. Rathgeb (30 min)

Quick UDP Internet Connections (QUIC) is a novel transport protocol introducing known features in a new protocol design. To investigate these features and the design, we developed a QUIC implementation in the INET simulation model suite. In this paper, we describe that implementation, its validation and a result achieved using the simulation model. The result shows the negative impact on throughput, when raising the acknowledgment ratio. We propose a solution and describe how it solves the issue.

12:40 - 14:00  Lunch Break

14:00 - 15:30Panel Discussion : On the Future of Internet Transport - click here to download the Panel Discussion PDF file

Panel Moderator: Gorry Fairhurst (Professor, University of Aberdeen)

15:30 - 15:45Coffee Break

15:45 - 17:15Session #3: Measurement Studies

Session Chair: Naeem Khademi

  • SABES: Statistical Available Bandwidth EStimation from passive TCP measurements - Francesco Ciaccia, Ivan Romero, Oriol Arcas-Abella, Diego Montero, René Serral-Gracià and Mario
    Nemirovsky (30 min)

Estimating available network resources is fundamental when adapting the sending rate both at the application and transport layer. Traditional approaches either rely on active probing techniques or iteratively adapting the average sending rate, as is the case for modern TCP congestion control algorithms. In this paper, we propose a statistical method based on the inter-packet arrival time analysis of TCP acknowledgments to estimate a path available bandwidth. Exhaustive experimentation on both simulations and real-world scenarios were conducted to validate our technique. Furthermore, we train an artificial neural network to improve the estimation accuracy.

  • On the Incompatibility of Scalable Congestion Controls over the Internet - Ferenc Fejes, Gergö Gombos, Sándor Laki and Szilveszter Nádas (30 min)

In addition to classic ones, a new family of congestion controls regarded as "scalable" has recently emerged. Scalable congestion control exploits ECN to provide much finer control as rate scales, enabling reactions proportional to the congestion level. Nowadays, different initiatives propose the use of Scalable congestion control over the public Internet, with the separation of Scalable and Classic flows. Similar to classic approaches, the evolution of Scalable congestion control has also begun as BBRv2 implemented a "DCTCP-inspired" ECN mechanism. In this paper, we investigate the compatibility of two available Scalable congestion controls, Data Center TCP, and Google's recent proposal called BBRv2, and evaluate them with different ECN marking strategies. Our evaluation relies on numerous measurements carried out in our testbed with 1 to 10Gbps bottleneck capacities and heterogeneous round trip times. At the bottleneck, the ECN marking of two traditional AQMs (RED and PI2) most commonly proposed for Scalable flows and a non-traditional core-stateless AQM (CSAQM) using innetwork resource sharing control is examined. After showing compatibility issues, we conclude that as long as the end-to-end flows control resource sharing, it will be tough to create a new, evolved congestion control that is fair to legacy solutions, and its deployment does not harm the Internet.

  • Evaluating QUIC’s Performance Against Performance Enhancing Proxy over Satellite Link - John Border, Bhavit Shah, Chi-Jiun Su and Rob Torres

Geosynchronous orbit satellite broadband serves a crucial role in bridging the digital divide by connecting under- served and unserved areas where terrestrial infrastructure is infeasible. Most of the TCP implementations are not optimized for high bandwidth delay product (BDP) satellite links. TCP Performance Enhancing Proxies (PEP) have alleviated degraded TCP performance in satellite broadband. The new transport protocol, QUIC, encrypts not only payloads but also most of the transport layer header. Consequently, it is no longer feasible to accelerate transport layer performance over a high BDP satellite link. The paper performs a measurement study to compare the performance of the QUIC protocol to that of the current accelerated transport layer over a GEO satellite link. The results showed that QUIC achieves only 20% of the throughput provided by the transport layer acceleration optimized for a satellite link. It also shows the need for an end- to-end loss recovery mechanism for QUIC.

17:15 - 17:30Concluding Remarks

Session Chairs: Naeem Khademi and Karl-Johan Grinnemo