The DataFlow Paradigm

  • Veljko Milutinović
  • Jakob Salom
  • Nemanja Trifunovic
  • Roberto Giorgi
Part of the Computer Communications and Networks book series (CCN)


This chapter presents an introduction to DataFlow supercomputing for big data problems. First, it explains why the DataFlow subject is becoming so important. More and more big data are present in all kinds of research or commercial challenges. Consequently, the DataFlow paradigm is getting importance, since it has been proven that it is the most suitable computing paradigm for big data. It offers superior speedups (depending on the application, from about 20 to about 200, even 2,000 in some isolated cases), as well as power savings (typically about 20 times); it brings the size reduction, too. A recent study by researchers of the Tsinghua University in China reveals that, for Shallow Water Weather Forecast (a big data problem), on the 1U level, compared to Tianhe1 (at the time of writing of this book, rated #1 on the Top500 Supercomputer List, which compares supercomputers based on Linpack, a small data benchmark), Maxeler (a DataFlow machine) demonstrates the speedup of 14. Second, it explains the hardware architecture, how the compiler works, and what the most suitable programming model is: programming in space. Third, it gives an overview of possible applications and the benefits to expect in all three domains of importance: speed, power, and size. Fourth, it tells about future expectations and how easy it is to use the DataFlow paradigm in the case of the Maxeler products: an example is given based on WebIDE (a Web-based integrated development environment).


FPGA Chip Chicago Mercantile Exchange Computing Equipment Common Reflection Surface DataFlow Programming 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. [Dennis1974]
    Dennis JB, Misunas DP (1974) A preliminary architecture for a basic data-flow processor. Newsl ACM SIGARCH Comput Archit News Homep 3(4):126–132CrossRefGoogle Scholar
  2. [DOE2014]
    US Department of Energy (2014) Advanced scientific computing research -X-Stack portfolio, April [Online]. Available:
  3. [Dongarra94]
    (2014) Top500 [Online]. Available:
  4. [Feynman96]
    Feynman PF (1996) Feynman lectures on computation. Addison-Wesley Publishing Company Inc., BostonGoogle Scholar
  5. [Flynn2013]
    Flynn M et al (2013) Moving from petaflops to petadata. Commun ACM 56(5):39–42. ACM, New York, NYGoogle Scholar
  6. [Gan2013]
    Gan L et al (2013) Accelerating solvers for global atmospheric equations through mixed-precision dataflow engine. In: Proceedings of the 23rd international conference on Field Programmable Logic and applications (FPL), Porto, Portugal, pp 1–6Google Scholar
  7. [Giorgi2014]
    Giorgi R et al (2014) TERAFLUX: harnessing dataflow in next generation teradevices. Microprocess Microsyst 38(8, Part B):976–990Google Scholar
  8. [Johnston2004]
    Johnston WM, Hanna JRP, Millar RJ (2004) Advances in dataflow programming languages. ACM Comput Surv 36(1):1–34CrossRefGoogle Scholar
  9. [Linux2000]
    Siever E et al (2009) Linux in a nutshell. O’Reilly Media, SebastopolGoogle Scholar
  10. [Maxeler2012]
    (2012) Exascale computing by the year 2018. Maxeler Technologies Ltd, LondonGoogle Scholar
  11. [Maxeler2014]
    (2014) The OpenSPL. Maxeler Technologies Ltd, LondonGoogle Scholar
  12. [Maxeler2015]
    (2015) Multiscale dataflow programming. Maxeler Technologies Ltd, LondonGoogle Scholar
  13. [Milutinovic88]
    Milutinovic V (1988) Computer architecture: concepts and systems. North-Holland, New YorkMATHGoogle Scholar
  14. [Milutinovic2014]
    (2014) BALCON: the gateway to ICT monitoring & control research in the Western Balkans, September [Online]. Available:
  15. [Moskowitz2014]
    Moskowitz H (2007) Selling blue elephant. Pearson Education Inc. Publishing as Prentice Hall, Upper Saddle RiverGoogle Scholar
  16. [Nemeth2008]
    Nemeth T et al (2008) An implementation of the acoustic wave equation on FPGAs. In: Proceedings of the 78th Society of Exploration Geophysicists (SEG) meeting, Las Vegas, November 2008, pp 2874–2878Google Scholar
  17. [SEG2008]
  18. [STFC2014]
    (2014) STFC Daresbury Laboratory first to install maximum performance computer (MPC), February 25 2014 [Online]. Available:
  19. [Stojanovic2013]
    Stojanovic S, Bojic D, Milutinovic V (2013) Solving Gross Pitaevskii equation using dataflow paradigm. IPSI Trans Internet Res Belgrade 9(2):19–22, SerbiaGoogle Scholar
  20. [Stojanovic2015]
    Stojanovic S, Milutinovic V (2015) A survey of dataflow architectures. Adv Comput Elsevier 96:1–45Google Scholar
  21. [WANG2010]
    Wang YH (2010) Multichannel matching pursuit for seismic trace decomposition, In Proceedings of the 72nd EAGE conference & exhibition incorporating SPE EUROPEC 2010, Barcelona, Spain, June 2010Google Scholar

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  • Veljko Milutinović
    • 1
  • Jakob Salom
    • 2
  • Nemanja Trifunovic
    • 3
  • Roberto Giorgi
    • 4
  1. 1.School of Electrical EngineeringUniversity of BelgradeBelgradeSerbia
  2. 2.MISANUBelgradeSerbia
  3. 3.Maxeler Technologies Inc.Palo AltoUSA
  4. 4.University of SienaSienaItaly

Personalised recommendations