SlideShare a Scribd company logo

Lightkone project : Lightweight computation for networks at the edge

•
•
Agence du Numérique (AdN)
Agence du Numérique (AdN)

The document summarizes the LightKone H2020 project, which aims to develop a model for general-purpose computing on edge networks. The project brings together 9 partners over 3 years with a budget of 3.57M€. It combines synchronization-free programming and hybrid gossip techniques to enable applications to run efficiently on heterogeneous edge networks despite their dynamic and unreliable nature. The goal is to address challenges of edge computing like latency, scalability, resilience and security by performing more computations locally at network edges rather than centralized data centers. Several use cases are highlighted like sensor-based management, content search, community network services and factory transportation. Open-source software is being developed to serve both independent edge and hybrid edge-datacenter models.

1 of 10
Download to read offline
November 6, 2017 PeterVan Roy LightKone coordinator Université catholique de Louvain TEKK Tour Digital Wallonia 2017 Mons International Congress Xperience (MICX) Project Lightweight computation for networks at the edge Horizon H2020 See Web site: lightkone.eu
Edge computing (Internet of Things) —  Edge computing is the next big frontier for the Internet ◦  In 2020 there will be 50 billion Internet of Things devices ◦  In 2027 there will be >1000 IoT devices per human on earth —  Killer apps are rapidly appearing for edge computing ◦  A good example is personal health monitoring: each human being will have a personal sensor array and a personal data center ◦  Data streams from the sensor array are continuously analyzed using deep learning algorithms to detect early signs of health problems ◦  This will reduce costs and increase quality of healthcare for all ◦  For example: C. Stamate et al,“Deep Learning Parkinson’s from Smartphone Data”, IEEE PerCom 2017, March 2017 LightKone H2020 project 2
—  LightKone: Lightweight computation for networks at the edge ◦  H2020 Project ID 732505, Jan. 1, 2017 – Dec. 31, 2019 ◦  9 partners (4 industrial, 5 academic), 3.57M€ budget —  The goal of LightKone is to develop a scientifically sound and industrially validated model for general-purpose computing on edge networks ◦  This is a challenge because edge networks are heterogeneous, loosely coupled, highly dynamic, and unreliable (frequent partitions, frequent offline or failed nodes) ◦  This is why today’s computations are mostly done on data centers! —  We take up the challenge because edge networks have great potential: ◦  Greatly increased scalability (edge devices far outnumber data center nodes) ◦  Much lower latency (edge computing makes real-time control possible) ◦  Increased resilience (the connection to the data center is a weak point) ◦  Increased security (edge computing avoids data center security and privacy issues) ◦  Local decision making ability (edge computing enables quick action in crisis situations) LightKone project in a nutshell LightKone H2020 project 3
Project partners 1.  UCL (Belgium) Université catholique de Louvain (coordinator) 2.  UPMC + INRIA (France) Université Pierre et Marie Curie + INRIA 3.  INESC TEC + UMinho Instituto de Engenharia de Sistemas e Computadores – (Portugal) Tecnologia e Ciência + Universidade de Minho 4.  TUKL (Germany) Technische Universität Kaiserslautern 5.  NOVA ID + UNL Associação para a Inovação e Desenvolvimento da FCT (Portugal) + Universidade Nova de Lisboa 6.  Scality (France) Scality 7.  Gluk (Netherlands) Gluk Advice BV 8.  UPC + Guifi (Spain) Universitat Politècnica de Catalunya (Guifi Community Network) 9.  Stritzinger (Germany) Peer Stritzinger GmbH Academic / industrial partners LightKone H2020 project 4
LightKone technology LightKone combines two recent advances in distributed computing to enable general-purpose computing on edge networks: —  Synchronization-free programming: Large-scale applications can run efficiently on edge networks by using convergent data structures (based on Lasp and Antidote from previous project SyncFree) → tolerates dynamicity and loose coupling of edge networks —  Hybrid gossip: Communication can be made highly resilient on edge networks by combining gossip with classical distributed algorithms (based on Plumtree epidemic broadcast trees used in industry) → combines naturally with synchronization-free programming LightKone H2020 project 5
Sensor-based management (Gluk) —  By adding edge computing, sensor and actuator arrays can manage systems directly on the edge —  We target precision agriculture, healthcare, and smart homes LightKone H2020 project 6
Advanced content search (Scality) —  Advanced search with indexed metadata (e.g., on image content) is done in data centers today —  Doing indexing directly at the edge gives much faster response and enables custom search based on machine learning LightKone H2020 project 7
Services for a community network (Guifi / UPC) —  Guifi.net is a common resource developed bottom-up by users —  It has more than 32,600 nodes (Jan. 2017), mostly in Catalonia —  We implement Guifi.net services by edge microclouds LightKone H2020 project 8 5/13 A few numbers about Guifi.net
Transport for manufacturing (Stritzinger) —  Automatic transport of partial products between processing stations in a factory —  Partial products have RFID tags and their paths through the factory are managed directly by the networked nodes LightKone H2020 project 9 NodeNodeNodeNode NodeNode Node Node
LightKone technology at your service —  We are developing open-source software both for light edge (edge by itself) and heavy edge (edge plus data center) ◦  Lasp edge computing platform for light edge: https://lasp-lang.org ◦  Antidote database for heavy edge: http://antidotedb.org —  Please contact us if you are interested in learning about our technology or collaborating ◦  https://lightkone.eu LightKone H2020 project 10

More Related Content

Lightkone project : Lightweight computation for networks at the edge

  • 1. November 6, 2017 PeterVan Roy LightKone coordinator Université catholique de Louvain TEKK Tour Digital Wallonia 2017 Mons International Congress Xperience (MICX) Project Lightweight computation for networks at the edge Horizon H2020 See Web site: lightkone.eu
  • 2. Edge computing (Internet of Things) —  Edge computing is the next big frontier for the Internet ◦  In 2020 there will be 50 billion Internet of Things devices ◦  In 2027 there will be >1000 IoT devices per human on earth —  Killer apps are rapidly appearing for edge computing ◦  A good example is personal health monitoring: each human being will have a personal sensor array and a personal data center ◦  Data streams from the sensor array are continuously analyzed using deep learning algorithms to detect early signs of health problems ◦  This will reduce costs and increase quality of healthcare for all ◦  For example: C. Stamate et al,“Deep Learning Parkinson’s from Smartphone Data”, IEEE PerCom 2017, March 2017 LightKone H2020 project 2
  • 3. —  LightKone: Lightweight computation for networks at the edge ◦  H2020 Project ID 732505, Jan. 1, 2017 – Dec. 31, 2019 ◦  9 partners (4 industrial, 5 academic), 3.57M€ budget —  The goal of LightKone is to develop a scientifically sound and industrially validated model for general-purpose computing on edge networks ◦  This is a challenge because edge networks are heterogeneous, loosely coupled, highly dynamic, and unreliable (frequent partitions, frequent offline or failed nodes) ◦  This is why today’s computations are mostly done on data centers! —  We take up the challenge because edge networks have great potential: ◦  Greatly increased scalability (edge devices far outnumber data center nodes) ◦  Much lower latency (edge computing makes real-time control possible) ◦  Increased resilience (the connection to the data center is a weak point) ◦  Increased security (edge computing avoids data center security and privacy issues) ◦  Local decision making ability (edge computing enables quick action in crisis situations) LightKone project in a nutshell LightKone H2020 project 3
  • 4. Project partners 1.  UCL (Belgium) Université catholique de Louvain (coordinator) 2.  UPMC + INRIA (France) Université Pierre et Marie Curie + INRIA 3.  INESC TEC + UMinho Instituto de Engenharia de Sistemas e Computadores – (Portugal) Tecnologia e Ciência + Universidade de Minho 4.  TUKL (Germany) Technische Universität Kaiserslautern 5.  NOVA ID + UNL Associação para a Inovação e Desenvolvimento da FCT (Portugal) + Universidade Nova de Lisboa 6.  Scality (France) Scality 7.  Gluk (Netherlands) Gluk Advice BV 8.  UPC + Guifi (Spain) Universitat Politècnica de Catalunya (Guifi Community Network) 9.  Stritzinger (Germany) Peer Stritzinger GmbH Academic / industrial partners LightKone H2020 project 4
  • 5. LightKone technology LightKone combines two recent advances in distributed computing to enable general-purpose computing on edge networks: —  Synchronization-free programming: Large-scale applications can run efficiently on edge networks by using convergent data structures (based on Lasp and Antidote from previous project SyncFree) → tolerates dynamicity and loose coupling of edge networks —  Hybrid gossip: Communication can be made highly resilient on edge networks by combining gossip with classical distributed algorithms (based on Plumtree epidemic broadcast trees used in industry) → combines naturally with synchronization-free programming LightKone H2020 project 5
  • 6. Sensor-based management (Gluk) —  By adding edge computing, sensor and actuator arrays can manage systems directly on the edge —  We target precision agriculture, healthcare, and smart homes LightKone H2020 project 6
  • 7. Advanced content search (Scality) —  Advanced search with indexed metadata (e.g., on image content) is done in data centers today —  Doing indexing directly at the edge gives much faster response and enables custom search based on machine learning LightKone H2020 project 7
  • 8. Services for a community network (Guifi / UPC) —  Guifi.net is a common resource developed bottom-up by users —  It has more than 32,600 nodes (Jan. 2017), mostly in Catalonia —  We implement Guifi.net services by edge microclouds LightKone H2020 project 8 5/13 A few numbers about Guifi.net
  • 9. Transport for manufacturing (Stritzinger) —  Automatic transport of partial products between processing stations in a factory —  Partial products have RFID tags and their paths through the factory are managed directly by the networked nodes LightKone H2020 project 9 NodeNodeNodeNode NodeNode Node Node
  • 10. LightKone technology at your service —  We are developing open-source software both for light edge (edge by itself) and heavy edge (edge plus data center) ◦  Lasp edge computing platform for light edge: https://lasp-lang.org ◦  Antidote database for heavy edge: http://antidotedb.org —  Please contact us if you are interested in learning about our technology or collaborating ◦  https://lightkone.eu LightKone H2020 project 10