Employment

RISE (formerly Swedish Institute of Computer Science) AB, 1995-present.

Telecommunications, automotive and machine learning for 20+ years. In the telecommunications sector, a large involvement with industry-led projects, a distributed file system called JetFile and some ST-II continuation.

Moved onto the “real-time Internet”, part of the ATM to IP migration, including kernel queuing algorithms (WFQ) and protocols such as RSVP. This led to my Ph.D topic, VoIP from several different aspects: dimensioning, modelling, buffer control, protocols (RTCP), measurements, admission control and quality assessment.

Caching, simulation. Modelling the caching behaviour of storage in the Internet. Implemented a model and tested with real traffic from Orange labs. As an ML scientist I analysed enormous data amounts namely Orange Labs’ whole French cellular HTTP network accesses. Caching strategies and mobility pattern searches the goals.

Radio network controller dimensioning task. Performance analysis and software package to systemise input, processing and suggestions with many thousands of parameters, with Ericsson. Recoded Excel into Visual Basic for easier error checking and version control.

New network architectures, caching and now Edge networking challenges. Two large EU projects looking at new structures for the Internet. Some design principles commonplace, e.g. the Pub-Sub paradigm.

Modelling 12 years of road traffic data on the E4 in Stockholm Sweden. Applied simple (fundamental diagrams of traffic flow) to complex AI models (LSTM). With Trafikverket.

Road traffic emissions estimation with LiU in a 3 year project with the Energy authority.

Co-developer of CheesePi measurement network. Python-based home measurement architecture for constant ‘always on’ monitoring.

Generic middleware for cyber physical systems, Based on WebAssembly, tested between Cloud-Edge with a real robot arm (TinkerKit Braccio). Large EU project (ANIARA) bringing together CPS, networking, AI and low power hardware for the network edge and cloud.

Sensor fusion of RGB, Thermal, LiDAR and RADAR data in a large EU project (‘ROADVIEW’).

IBM, Germany GmBH, 1993-1995 (2 years).

Worked with developing streaming protocols for the Internet. Now commonplace, but with SICS, and BBN in the USA we developed a protocol called ST-II, the ‘IPV5’ of its time. From IBM’s device driver to webapps we coded a complete protocol suite, called HeiTS, as part of Germany’s capital move Bonn→Berlin. Coded an interactive kiosk banking application in OS/2.

CSIRO, Australia, 1990-1992 (2 years).

Collaborated closely with Australian-based industries. Worked on a commercial statistical package which is now R. We developed new routines and its’ first graphical interface.

3D Pixel Labs, UK (now nVidia), 1988-1990 (2 years).

Microcoded a 3d graphics engine, based on a bitslice architecture, combining both pixel and vector graphics. Extended the bitslice architecture to 4×4 & 16×16 SIMD architectures, together with a specialised graphics language called Pixel C.

Education 

Ph.D. Electrical Engineering, KTH, Sweden, 2009.
M.Sc. Computer Science, Manchester, UK, 1988.
B.Sc.  Physics, Manchester, UK, 1987.

Publications

  • I. Marsh, The Responsiveness and Deployment of WebAssembly in Cyber-Physical Systems, submitted to EdgeSys 2024.
  • John, W. et.al , ANIARA Project – Automation of Network Edge Infrastructure and Applications with Artificial Intelligence, 2023
  • I. Marsh et al. Evolving 5G: ANIARA, an Edge-Cloud perspective, Computing Frontiers, 2021.
  • I. Marsh, Dimensionality reduction in (large) measurement datasets, 2020.
  • B. Bjurling, I. Marsh, Per Kreuger, Data Readiness in the BADA project, technical report, 2017.
  • L. McNamara, I. Marsh, S. Forslin, CheesePi: A Raspberry Pi based measurement platform, Research and Applications of Internet Measurements,  RAIM, 2015.
  • I. Marsh, VANET communication: a traffic flow approach, PIMRC 2012.
  • A. Gunnar and I. Marsh, Efficient cache management for content distribution in Information Centric Networks, SNCNW, 2012.
  • I. Marsh, F. Li and G. Karlsson, Wide Area Measurements of VoIP Quality, QoIFS, 2003.
  • O. Hagsand, I. Marsh and K. Hanson, Sicsophone: A Low-delay Internet Telephony Tool, Euromicro, 2003.
  • H. Abrahamsson, O. Hagsand and I. Marsh, TCP over Variable Capacity Links: A Simulation Study, WfPHSN, 2002.
  • B. Ahlgren, A. Andersson, O. Hagsand and I. Marsh, Dimensioning Links for IP Telephony, IPTEL 2001.
  • I. Marsh, Measuring Internet Telephony Quality: Where are we today?, Globecom 1999.
  • B. Grönvall, S. Pink, I. Marsh, JetFile: a multicast based file system for the Internet, SIGOPS 1996.

Program committees

  1. EU project reviewer, 2020-
  2. TMA technical program committee member, 2019.
  3. VTC program committee member, 2018.
  4. Big DAMA program committee member, 2016.
  5. QoS, Reliability and Performance Modelling Graz, Austria, 2008.
  6. 13th Multimedia Computing and Networking, 2006.
  7. Next Generation Teletraffic and Wired/Wireless Advanced Networking, 2004/5.

Professional and academic experience

  1. Program committee of Computer Networks, Elsevier, 2005-present.
  2. Program committee of Computer Communications, Elsevier, 2005-present.
  3. Co-adviser of 3 Ph.D students in Porto, 2010.
  4. Supervised 20 masters students at KTH, 2001-present.
  5. Supervised six undergraduates in a systems design course at KTH. 
  6. Main reviewer for Addison Wesley’s IP telephony books.
  7. Reviewed over 150 articles for conferences and journals.
  8. Member of IEEE Comsoc and Information theory societies.

IBM, Germany GmBH, 1993-1995.

Worked with developing streaming protocols for the Internet. Now commonplace, but with SICS, and BBN in the USA we developed a protocol called ST-II, the ‘IPV5’ of its time. From IBM’s device driver to webapps we coded a complete protocol suite, called ‘HeiTS’, it was part of Germany’s capital move Bonn→Berlin. Coded an interactive kiosk banking application in AIX and OS/2.

Also worked with a distributed file system AFS, which with ST-II were the first tasks at SICS (above).

CSIRO, Australia, 1990-1992.

Collaborated closely with Australian-based industries. Worked on a commercial statistical package which is now, the publically available package, R. We developed new routines and its’ first graphical interface.

3D Pixel Labs, UK (now nVidia), 1989-1990.

Microcoded a 3d graphics engine, based on a bitslice architecture, combining both pixel and vector graphics. We later extended the bitslice architecture to 4×4 and 16×16 SIMD architectures, together with a specialised graphics language called Pixel C.

Funding efforts

READY (Co-PI, SSF project) ANIARA (11M Euro), TENS (5MEuro), ROADVIEW (11M Euro).

Education 

Ph.D. Electrical Engineering, KTH, Sweden, 2009.
M.Sc. Computer Science, Manchester, UK, 1988.
B.Sc.  Physics, Manchester, UK, 1987.

Program committees

  1. TMA technical program committee member, 2019.
  2. VTC program committee member, 2018.
  3. Big DAMA program committee member, 2016.
  4. QoS, Reliability and Performance Modelling Graz, Austria, 2008.
  5. 13th Multimedia Computing and Networking, 2006.
  6. Next Generation Teletraffic and Wired/Wireless Advanced Networking, 2004/5.

Professional and academic experience

  1. Program committee of Computer Networks, Elsevier, 2005-present.
  2. Program committee of Computer Communications, Elsevier, 2005-present.
  3. Co-adviser of 3 PhD students in Porto, 2010.
  4. University of Oslo, evaluator of a multimedia course 2007.
  5. Supervised 15 masters students at KTH, 2001-present.
  6. Supervised six undergraduates in a systems design course at KTH. 
  7. Main contact point for Ph.D students whilst at SICS.
  8. Main reviewer for Addison Wesley’s IP telephony books.
  9. Reviewed over 150 articles for conferences and journals.
  10. Member of IEEE Comsoc and Information theory societies.

Major technical accomplishments

1. Traffic flow prediction for Stockholm’s highways (2018). Using historical, and real time traffic flow data, we have trained a Deep Neural network to predict traffic flows, up to 30 mins in advance. We used a special Deep network called LSTM (Long-Short-Term-Memory), which captures both short (i.e. rapid braking) and longer aspects (filling up to capacity) on the road. A paper European Conference on Data Analysis, demonstration and a Python software package based on Keras is available.

2. Measurements in Sweden’s 3/4G networks using 4 operators (2018). With the use of an EU FIRE+ platform called MONROE, we measured and compared the performance of operators in busses within rural Sweden. We compared the performance using different types of TCP, UDP as well as choosing the best interface given radio information (e.g. RSSI). We conclude that the best operators vary along a buss’s path, and having more than 1 interface (and operator) improves up and downlink performance.

3. Co-creator of CheesePI network measurement platform (ongoing). CheesePI is an Internet measurement infrastructure using Raspberry PIs targeted towards Quality of Experience for the home user. It has been developed at SICS part of Swedish-ICT. CheesePI measures the Internet loss, delay, jitter, as well as the number of hops between PIs placed at selected locations and well-known media sites. The goal of the project is to measure the quality in a free, neutral and unbiased manner. Individual PIs store the measurement data in a schemaless database, whilst the central server (cheesepi.sics.se) stores the aggregated data from each PI for analysis, visualisation and dissemination. Each user can see their historical home quality using a networked device (phone, computer, TV) via a customisable dashboard. Networking tools can be added and removed without major rewrites, as the authors envisage a community- based approach to network monitoring at an affordable price. The PIs send periodic data to probe each connection, particularly during popular events, to ascertain the perceived home quality experience. All the source code is available and free to download.