Karl-Johan Grinnemo

My research primarily considers transport-level solutions for IP-based applications with soft real-time requirements in both data communication and telecommunication networks. Currently, I am involved in two research projects, an EU COST action, and a Knowledge Foundation research profile: A six year project, ”Research Environment for Advancing Low Latency Internet” (READY); a three-year project, ”A New, Evolutive API and Transport-Layer Architecture for the Internet” (NEAT); the ICT COST action IC1304, ”Autonomous Control for Reliable Internet of Services (ACROSS); and the research profile, ”High Quality Networked Services in a Mobile World” (HITS).

The READY project aims to ameliorate the latency characteristics of IP applications in general, but with a particular focus on applications in the media, telecom, and industrial automation industries. Latency has in recent years become a key issue for the continued success of Internet, and its expansion into novel application areas such as cloud services. My research involves finding ways of using multiple network paths and Multipath TCP (MPTCP) to reduce the latency between cloud data centers and end-user mobile devices. An initial study, in which traffic from three typical cloud applications were studied, Netflix, Google Maps, and Google Docs, suggested that significant latency reductions could indeed be possible through the use of MPTCP, and then particularly for high-intensity video traffic such as Netflix. Still, MPTCP only to some extent addresses latency, and thus I find it reasonable to think that further latency reductions for latency-sensitive flows are feasible provided MPTCP is extended with latency-aware path management and/or -selection schemes.

The NEAT project addresses the fact that it has become extremely difficult to introduce novel transport services on the Internet, not least transport services for latency-sensitive Web applications: NATs, firewalls, and various other middleboxes, tend to filter out all traffic that is not TCP or UDP. Rather than trying to find an overall solution to this problem, previous and current work have primarily considered technical compromises and workarounds. For example, HTTP has become an application transport protocol for many Web applications, e.g., JSON, SOAP, RSS, and HTTP Live Streaming. Furthermore, new application-layer transport protocols that use the current underlying transport layer as a communication substrate have been proposed, e.g., QUIC and Minion. In contrast, NEAT argues for a more complete solution to the ossification problem – a transport system that decouples applications from the choice of actual transport protocol being used, and that enables applications to explicitly communicate their service requirements via a new transport API. My work in the project includes participating in the design of the transport system; design mechanisms to dynamically control data transmission based on policies and resource tradeoffs; and, design experimental transport protocol mechanisms that let the NEAT transport system offer a larger spectrum of transport services. As part of the latter work, we intend to introduce latency-aware MPTCP solutions from the READY project into NEAT.

In the EU COST action, Across, Karlstad University together with Ericsson and the University of the Basque Country are working towards congestion control schemes that provide low delay and high throughput in current and future cellular networks. The rationale behind this work is to propose transport protocols that meet the quality of service requirements of a future Internet of Services, i.e., an Internet where everything that is needed to use software applications is available as a service on the Internet, e.g., the software itself, the tools to develop the software, and the platform to run the software.

As part of our research profile, HITS, scalable and cost-efficient communication solutions in future telephony signaling systems are considered. Together with Ericsson, we have conducted an experiment with the overarching aim to determine if the use of the transport protocol SCTP, with extensions for concurrent multipath transfer (CMT-SCTP), could provide a faster startup behavior for telephony signaling traffic than standard SCTP. The results from the experiment were indeed encouraging, and suggested that CMT-SCTP in several cases could give a faster startup behavior than standard SCTP over a multipath association – in spite of some of the paths having a packet-loss rate of several percent. On the basis of this, we are currently studying various schemes for CMT-SCTP to schedule packets over the available paths, to reduce latency, while at the same time upholding a high throughput and bandwidth utilization.

Karlstad University

• Course responsible and examiner for undergraduate course Distributed Systems and Applications  (DVGC15), Fall 2010 – onwards.
• Course responsible and examiner for undergraduate course Computer Engineering Project  (DVAE08), Fall 2015 - onwards.
• Lecturer in graduate course Perspectives in Computer Science  (DVAD07), Fall 2011 – Fall 2015. Assistant in Fall 2016.
• Assistant in Ph.D. course Individual Themes , Spring 2014 and Fall 2015.
• Assistant in Ph.D. course Classic Papers , Fall 2013.
• Lab assistant in undergraduate course Programming Techniques for Engineers  (DVGA15), Fall 2013.
• Lab assistant in undergraduate course Software Development and Design  (DVGA09), Spring 2013.
• Lab assistant in undergraduate course Operating Systems  (DVGB01), Fall 2012 and Fall 2013.
• Thesis advisor in undergraduate course Bachelor’s Project  (DVGC25), Fall 2011 – Fall 2013.
• Lecturer in Ph.D. course Current Advances in Computer Networking, Spring 2011.

KTH Royal Institute of Technology

• Project assistant in graduate course "Communication System Design" (IK2200), Spring 2010.
• Course responsible for undergraduate course "Networks and Communication" (IK1203), Spring 2010.
• Lecturer and lab assistant in graduate course "Advanced Internetworking" (IK2215), Fall 2009 and Fall 2010.
• Lecturer and lab assistant in graduate course "Internet Security and Privacy" (IK1206), Fall 2009.

• Cooperates with SICS Swedish ICT AB and Mälardalen University in the READY research project.
• Cooperates with Ericsson AB in the KKS HITS project.
• Cooperates with University of Oslo, Simula Research Laboratory, University of Aberdeen, FH Münster, Celerway, Mozilla, and EMC in the EU HORIZON 2020 NEAT project.
• Guest lecturer at KTH.

I completed my Ph.D. studies in June 2006, and worked half a year as a post doctor at the former Department of Computer Science at Karlstad University. During that time, I continued my research from my Ph.D. studies on telephone signaling over IP and SCTP together with Ericsson Research in Aachen. I was also involved in a cooperative research effort on using SCTP in mobile networks together with Universitat Politècnica de Catalunya (UPC) in Barcelona. In January 2007, my post doctoral position at Karlstad University ended, and I resumed my position as consultant at Tieto in Karlstad. My work at Tieto was in many ways aligned with my research, and covered a broad range of activities including pre-studies, system design, and implementation. For example, I was involved in the implementation of an alarm function in Ericsson’s Connectivity Packet Platform (CPP), and conducted a pre-study together with colleagues from Tieto’s Beijing office on how to build an SCTP traffic offload engine for SCTP in CPP. Moreover, I continued to have a foot in research. As an adjunct researcher, I was actively participating in an on-going research effort between Tieto and Karlstad University. I took part in more or less all research project phases including planning, grant seeking, and execution.

Between the fall of 2009 and the fall of 2010, I was on leave from Tieto in Karlstad and worked as an acting associate professor at the School of Information and Communication Technology at the KTH Royal Institute of Technology. As an associate professor, I initiated a research project, ”Smart Session Management for Multi-homed SCTP” (SCTP Smartswitch), together with Karlstad University and Ericsson Research in Kista. The “SCTP Smartswitch” project aimed at extending the failover/changeover mechanism of SCTP with smart session management so that it would be able to manage everything from real-time applications, such as interactive games, which require fast failovers/changeovers, to roaming mobile terminals which sometimes experience long disconnection periods, i.e., long failover/changeover periods. In the fall of 2010, I acquired a four-year, tenure-track position at Karlstad University, and during the first two years as associate senior lecturer at their computer science department, the ”SCTP SmartSwitch” project was a major part of my research work. Notably, it resulted in about ten publications, an SCTP-based mobility management framework for smartphones and tablets that offered sub-second vertical handover between 3G and WiFi networks, and a generic SCTP-based session layer for mobility support in mobile delay-tolerant networks.

I became a Senior Lecturer at Karlstad University in the Fall of 2014. In recent years, my research has to a large degree focused on the use of multi-path transport protocols such as Multipath TCP and CMT-SCTP to increase reliability and throughput and decrease latency in IP networks. I have authored and co-authored around forty conference and journal papers, and is a Senior member of IEEE.