IEEE International Conference on High Performance Switching and Routing
18-21 June 2017 – Campinas, Brazil

Keynote Talks

    Christian Esteve Rothenberg  University of Campinas, Brazil
    Gergely Pongracz  Senior specialist in Ericsson Research in the area of complex packet processing
    Lena Wosinska  KTH Royal Institute of Technology, Sweden
    Luiz A. Dasilva  Trinity College Dublin, Ireland
    Sergio Celaschi  CTI Renato Archer, Campinas, Brasil


Keynote Talk 1: Softwarized Dataplanes and the P^3 trade-offs: Programmability, Performance, Portabiilty

Christian Esteve Rothenberg (University of Campinas, Brazil)

Christian Esteve Rothenberg is an Assistant Professor in the Faculty of Electrical & Computer Engineering (FEEC) at University of Campinas (UNICAMP), Brazil, where he received his Ph.D. (2010) and currently leads the Information & Networking Technologies Research & Innovation Group (INTRIG). His research activities span all layers of distributed systems and network architectures and are often carried in collaboration with academia and industry (e.g., Ericsson, Samsung, CPqD) around the world, leading to multiple open source networking projects (e.g., RouteFlow, libfluid, ofsoftswitch13, Mininet-WiFi) in the areas of SDN and NFV among other scientific results. Christian has two international patents and over 70 publications in scientific journals and top-tier networking conferences including SIGCOMM and INFOCOM. Christian worked as Senior Research Scientist in the areas of IP systems and networking at CPqD R&D Center (2010-2013), and holds the Telecommunication Engineering degree from the Technical University of Madrid (ETSIT – UPM), Spain, and the M.Sc. (Dipl. Ing.) degree in Electrical Engineering and Information Technology from the Darmstadt University of Technology (TUD), Germany, 2006.

Summary: The realization of network softwarization, an overarching buzzword to encompass all software-centric developments from the Software-Defined Networking (SDN) and Network Function Virtualization (NFV) trends, is being enabled through a set of innovations in high-speed data plane design and implementation. Recent efforts include te-architecting the hardware-software interfaces and exposing programmatic interfaces (e.g., OpenFlow), programmable hardware-based pipelines (e.g. Protocol Independent Switch Architecture – PISA) along suitabe programming languages (e.g., P4), and multiple advances on low overhead virtualization and fast packet processing libraries (e.g. DPDK, for Linux based general purpose processor platforms. This talk provides an overview of relevant ongoing work and discusses the trade-offs of each design and implementation choice of software-defined dataplanes regarding Programmability, Performance, and Portability.

Keynote Talk 2: How data plane programmability can help in fully utilizing 5G capabilities

Gergely Pongracz (Senior specialist in Ericsson Research in the area of complex packet processing)

2016: Leader of the DP/NFV team in the 5G Realization Program in Ericsson Research. This program ended with well received demos at the ER Days and the Mobile World Congress 2017 in Barcelona. Led university projects (Hungary, Brasil) which began to show good results: results were showed at one of the highest quality networking conference, IEEE SigComm (both OpenFlow – paper + demo and P4 work – 2 demos).
– 2015: Presented the “Virtual switching innovation” proof point at the Mobile World Congress in Barcelona which was followed by other presentations and tech-transfer activities. Started and supervised two university projects: the data plane project with BME produced huge amount of performance measurement results which later were used for high quality publications and also Ericsson internal communication, while the P4 activity at ELTE started a bit slower showing the first major results in 2016.
– 2014: Major contributor in the EU FP7 Unify project lead by ETH (WP5 Universal Node – data plane) + leader of an internal SDN project involving also people from India (DUIB).
– 2013: This year was about learning new technologies and gaining knowledge in high-performance packet processing both in terms of software (e.g. DPDK) and hardware (e.g. CPUs/NPUs). Results presented e.g. at SigComm HotSDN 2013
– 2019-12: Working in the area of deep packet inspection contributing in several studies and PoC prototypes

Patents and publications:
– authored or co-authored 26 filed patent applications
– recent publications:
o Linaro Connect 2017 (invited talk) – P4 over ODP: solution, results and comparison to a different approach
o SigComm 2016 – Dataplane Specialization for High-performance OpenFlow Software Switching
o EWSDN 2015 – Cross-Platform Estimation of Network Function Performance
o IEEE HPSR 2015 (invited paper) – Universal Node: Towards a High-Performance NFV Environment
o SigComm HotSDN 2013 – Cheap Silicon: a Myth or Reality? Picking the Right Data Plane Hardware for Software Defined Networking
o EWSDN 2013 – Removing Roadblocks from SDN: OpenFlow Software Switch Performance on Intel DPDK

External relations:
– strong connections to Intel’s DPDK and OVS-DPDK teams
– building up also good connections with Linaro’s ODP (Open Data Plane) group and ARM ecosystem players (e.g. Cavium, Freescale)
– extensive relationships with the Budapest University of Technology and Economics, ELTE (Science University in Budapest), Unicamp (Brazil) and most recently Berkeley (USA)

Summary: In the era of 5G network and application developers has to fight with two, seemingly contradictory
requirements. While network performance should be increased by an order of magnitude both in
terms of delay, bandwidth and connected devices, the flexibility of the infrastructure must be increased to allow rapid feature introduction, easy error tracking and good separation of concerns
and as a result allow the same innovation possibilities in the networking area as it is possible in IT.

In the coming years the network has to become an enabler rather than a limitation. When we are
talking about 5G, we are talking about a wide range of services from traditional mobile broadband
including bandwidth intensive downloads and high-quality interactive video talks to (currently) sci-
fi like possible applications like coordinated remote driving, wireless controlled exoskeletons, or
automated factories with wireless connected robots.

So instead of one or two dominating applications like it was the case in 3G, in the 5G network we
have to prepare for a large number of smaller applications with very different requirements. This
means that it is economically impossible to build physically separated networks which means we
have to have good ways of fulfilling the requirements even if there are several different applications sharing the same physical network. So in the future resource sharing and virtualization will be a key enabler from both technological and economic aspects.

But NFV is with us for 4-5 years now and it seems the takeoff is much slower then it was expected when the hype started. There are several possible issues, some of which could even be show stoppers: e.g. large performance overhead, lack of a stable and functionality-wise good-enough execution environment and lack of interoperability between NFV applications and execution environments. Is there a natural tradeoff between high-performance and good programmability? How a domain-specific language, like P4 can help in this problem area, if it can at all? How a future NFV system should look like, and what is the boundary between access and the cloud if there is such boundary at all? Several questions that inspire lots of new and interesting research work around the globe recently.

In the presentation I’ll show what are the main drivers behind the changes that are happening these days and what could we expect to see when 5G comes into reality. After showing a few use cases that would be enabled by a well-designed 5G network and the higher level concepts such as the components and platform features of a multi-service network I’ll try to argue that the increased complexity in packet processing requires solutions like P4 by going thru details of our P4 work touching both functionality and performance aspects.

Our P4 research has been so far focusing on creating an optimal compiler / pipelines on different CPU-based, programmable platforms, such as Intel x86 and ARM, so most of the results are related to these, but the presentation will also contain real-life results obtained with other software pipelines, such as (extended) OpenFlow software switches with NFV capabilities and also expected results on hardware platforms such as NPUs and switch ASICs based on our modelling work.

Keynote Talk 3: Optical Datacenter Networks: Challenges and Recent Developments

Lena Wosinska and Jiajia Chen (KTH Royal Institute of Technology, Sweden)

Lena Wosinska received her Ph.D. degree in Photonics and Docent degree in Optical Networking from the KTH Royal Institute of Technology in Stockholm, Sweden, where she is currently a Full Professor in Telecommunication in the School of Information and Communication Technology (ICT). She is founder and leader of the Optical Networks Lab (ONLab) at KTH/ICT. She has been working in several EU projects and coordinating a number of national and international research projects.

Her research interests include fiber access networks, energy efficient optical networks, photonics in switching, optical network control, reliability and survivability, and optical datacenter networks. She has been involved in many professional activities including guest editorship of IEEE, OSA, Elsevier and Springer journals, serving as General Chair and Co-Chair of several IEEE, OSA and SPIE conferences and workshops, serving in TPC of many conferences, as well as being reviewer for scientific journals and project proposals. In 2007–2009 she was an Associate Editor of OSA Journal of Optical Networking, and in 2009 – 2013 she has been Associate Editor of IEEE/OSA Journal of Optical Communications and Networking. Currently she is serving on the Editorial Board of Springer Photonic Networks Communication Journal

Summary: Due to the growing importance of IT services and dramatically increasing traffic volume, interconnection networks within datacenters are facing serious challenges in terms of capacity, energy consumption, flexibility, reliability and latency. This talk will address these challenges and report for recent developments to achieve

– ultra-high capacity inside the datacentre, particularly for large-scale site infrastructure that hosts more than several hundred thousand servers,

– better energy-efficiency throughout all parts of the interconnection network within the datacenter,

– flexible provisioning of connectivity inside the datacenter in order to efficiently utilize network resources and support differentiated services, and

– high reliability and low latency in order to support emerging critical services, such as connected vehicles, tactile Internet.

It is no doubt that photonic technology will become the key option to address the challenge of ultra-high capacity and scalability of interconnection networks of the large-scale datacenters. However, how to best take advantage of optical transport and develop highly scalable, energy-efficient, flexible, reliable and low-latency solutions becomes a crucial challenge. In this talk, we will present recent developments in optical datacenter network architectures and technologies. We will also highlight open questions and problems to be solved in order to meet the future demand.

Keynote Talk 4: 5G: Adaptable Networks Enabled by Versatile Radio Access Technologies

Luiz A. DaSilva (Trinity College Dublin)

Luiz A. DaSilva holds the chair of Telecommunications at Trinity College Dublin, where he is a co-principal investigator of CONNECT, a telecommunications centre funded by the Science Foundation Ireland. Prior to joining Trinity College, Prof DaSilva was with the Bradley Department of Electrical and Computer Engineering at Virginia Tech for 16 years. His research focuses on distributed and adaptive resource management in wireless networks, and in particular radio resource sharing and the application of game theory to wireless networks. Prof DaSilva is a principal investigator on research projects funded by the National Science Foundation, the Science Foundation Ireland, and the European Commission. Prof DaSilva is an IEEE Communications Society Distinguished Lecturer and a Fellow of the IEEE, for contributions to cognitive networks and to resource management in wireless networks.

Summary: The diverse and demanding requirements for the next generation of mobile networks necessitate a shift away from the rigid networks of previous generations, towards greater versatility and adaptability. Essential enablers for this versatility include: new, more flexible radio access technologies; the ability to share network infrastructure and spectrum; and dynamic management of resources across the wireless access and the optical backbone and core networks. In this presentation, we will discuss how some wireless technologies being considered for 5G can also bring a higher degree of flexibility to the managament of network resources. We will also present some recent results on the performance of shared networks, and introduce the Horizon 2020 project FUTEBOL, which is setting up research infrastructures in Brazil and Europe to enable experimentation at the convergence point between optical and wireless networks.

Keynote Talk 5: Networking for Prevention and Monitoring of Natural Disasters

Sergio Celaschi (CTI Renato Archer, Campinas, Brasil)

Sergio Celaschi was born in the city of Sao Paulo – Brazil in 1951. He graduated in Applied Physics from Stanford University in 1985. From 1985 to 2000 he joined CPqD – The Center for Research and Development of TELEBRAS, working on the technology of optical fibers and fiber optics devices and systems. He taught and researched at several Brazilian universities. He currently works at the Information Technology Center CTI Renato Archer, a R&D unit of the Brazilian Ministry of Science, Technology, Innovation and Communication (MCTIC). He has 30+ years of experience in technological developments and technology transfer activities. His technical contributions cover the areas of: industrial automation; digital image processing; machine vision; RFID; optical network architecture; fiber communication and optical devices manufacturing and technology transfer; biomedical engineering; biophysics, and applied physics. He developed processes and products in areas of low temperature physics, telecommunications engineering, industrial automation, machine vision, and laser therapy for health. Has recent experience in the field of climate change and remote monitoring of meteorological data.

Summary: A report on the status of the Brazilian public network of automatic hydro-meteorological stations will be presented. This network was recently projected, acquired, installed, and is maintained by the “Centro de Monitoramento e Alerta de Desastres Naturais – CEMADEN”, with the aid of CTI Renato Archer and Facti – Fundação de Apoio a Capacitação em Tecnologia da Informação, to provide a public system for monitoring natural disasters. CEMADEN provides early warnings for natural disasters affecting Brazil, analyses and issues alerts associated to floods and impacts of severe droughts. The challenges of planning, installing and maintaining such a network on a continental area comprising diversified environments and distinct socioeconomically scenarios will be discussed.