Jakob Salom

Guide to DataFlow Supercomputing: Basic Concepts, Case Studies, and a Detailed Example

This unique text/reference describes an exciting and novel approach to supercomputing in the DataFlow paradigm. The major advantages and applications of this approach are clearly described, and a detailed explanation of the programming model is provided using simple yet effective examples. The work is developed from a series of lecture courses taught by the authors in more than 40 universities across more than 20 countries, and from research carried out by Maxeler Technologies, Inc. Topics and features: presents a thorough introduction to DataFlow supercomputing for big data problems; reviews the latest research on the DataFlow architecture and its applications; introduces a new method for the rapid handling of real-world challenges involving large datasets; provides a case study on the use of the new approach to accelerate the Cooley-Tukey algorithm on a DataFlow machine; includes a step-by-step guide to the web-based integrated development environment WebIDE.

New benchmarking methodology and programming model for big data processing

Big data processing is becoming a reality in numerous real-world applications. With the emergence of new data intensive technologies and increasing amounts of data, new computing concepts are needed. The integration of big data producing technologies, such as wireless sensor networks, Internet of Things, and cloud computing, into cyber-physical systems is reducing the available time to find the appropriate solutions. This paper presents one possible solution for the coming exascale big data processing: a data flow computing concept. The performance of data flow systems that are processing big data should not be measured with the measures defined for the prevailing control flow systems. A new benchmarking methodology is proposed, which integrates the performance issues of speed, area, and power needed to execute the task. The computer ranking would look different if the new benchmarking methodologies were used; data flow systems would outperform control flow systems. This statement is backed by the recent results gained from implementations of specialized algorithms and applications in data flow systems. They show considerable factors of speedup, space savings, and power reductions regarding the implementations of the same in control flow computers. In our view, the next step of data flow computing development should be a move from specialized to more general algorithms and applications.

Paradigm Shift in Big Data SuperComputing: DataFlow vs. ControlFlow

The paper discusses the shift in the computing paradigm and the programming model for Big Data problems and applications. We compare DataFlow and ControlFlow programming models through their quantity and quality aspects. Big Data problems and applications that are suitable for implementation on DataFlow computers should not be measured using the same measures as ControlFlow computers. We propose a new methodology for benchmarking, which takes into account not only the execution time, but also the power and space, needed to complete the task. Recent research shows that if the TOP500 ranking was based on the new performance measures, DataFlow machines would outperform ControlFlow machines. To support the above claims, we present eight recent implementations of various algorithms using the DataFlow paradigm, which show considerable speed-ups, power reductions and space savings over their implementation using the ControlFlow paradigm.

An Example Application: Fourier Transform

This chapter represents an example of accelerating the Cooley-Tukey algorithm with the Maxeler MAX3 machine and gives the results of the achieved acceleration. First, it explains the importance and usages of the Cooley-Tukey algorithm. Second, it gives mathematical explanation of the algorithm and algorithm’s pseudo code and explains different ways to implement the algorithm. The implementation with best time and memory complexity is explained in detail. Third, it explains how the algorithm has been accelerated using DataFlow engines. Fourth, it explains the experiments done to measure acceleration and present the results. The final results are presented as various graphs with explanations.

Transforming Applications from the Control Flow to the Dataflow Paradigm

This chapter analyzes potentials of accelerating applications by transforming them from control flow to dataflow representation and mapping them directly to the hardware based on the FPGA. Firstly, potentials for improvements will be analyzed. Both reduction in execution time and power consumption will be analyzed. Transforming control flow to dataflow applications will be analyzed on a Huxley muscle model implemented using the dataflow approach.

Maxeler AppGallery Revisited

The reason to include a chapter with this content into this all-around Maxeler book was to show the wide audience, the audience that is still not familiar with dataflow programming paradigm, that there is help and that there is a whole library of tools and applications that can help a programmer in times of need. There are so many detailed source codes and many tools that can either be analyzed and easily changed or used as such, in new applications. Another reason for describing the items in the gallery is to present to the wider audience a good plethora of different cases Maxeler dataflow architecture can be used at. Next to each AppGallery item, there is a short description and applicable use of cases. Not all of the items are described. Some of them are skipped because they are differently and more in depth presented in Trifunovic et al. (J Big Data 3:4, 2016. Available at: https://journalofbigdata.springeropen.com/articles/10.1186/s40537-015-0038-8) and some are omitted because their usage possibilities are too limited.

Discrepancy Reduction Between the Topology of Dataflow Graph and the Topology of FPGA Structure

One of the main limiting factors in dataflow supercomputing is the discrepancy between the topology of a typical dataflow graph (produced by compiler) and the typical topology of FPGA structure (produced by manufacturer) onto which the execution graph has to be mapped. One possible school of thought is to study cases where infinitesimal changes in hardware domain may generate much more than infinitesimal impact in the benefit domain (speed/complexity/power/risks). The research analyzes the effects of one such infinitesimal add-on in the hardware domain (moving from two-input adders to three-input adders). Different compilation techniques, debugging and optimizing tools and methods, offered by “Maxeler Technologies Ltd.,” are presented and elaborated.

Polynomial and Rational Functions

Nowadays, when technology and science develop fast, the speed of data processing has become very important. Also, with science developing so broadly, there is an increase in the amount of data being processed for solving complex systems in physics, chemistry, biology, astronomy, seismology, meteorology, medicine, etc. For example, in meteorology, which is a science about the Earth’s atmosphere and the changes that happen within it, some of the main occurrences that are being studied are the amount and kind of precipitation, thunderstorms, tornadoes, tropical cyclones, and typhoons; in order to make the weather forecast as precise as possible, it is necessary to collect large amounts of data whose processing is based on solving dynamic equations as functions of time. It is very important to process large scales of information in a short time and give prediction of possible bad weather conditions in order to save people lives and property in a certain area. This chapter deals with resolving this kinds of problems, that is, it describes principle that is used to process large amounts of data in a short period of time. It will be described on an example of calculating values of a polynomial and rational function on the Maxeler architecture. Polynomials are used in engineering for designing roads, buildings, etc. In economy they are used for modeling patterns of economical growth and in medicine for describing bacterial behavior.

The Infrastructure Framework for Mind Genomics

This chapter links Mind Genomics with various techniques used to create electronic stores for electronic business on the Internet. Since Mind Genomics could exist only in the Internet-based environments, it is treated as an add-on on the top of some appropriate e-Business infrastructure. Notes were made in relation to three different pioneering techniques used for creation of electronic stores on the Internet, and the readers are pointed to more detailed sources, in a previous textbook of one of the authors of this textbook.

Introduction to Basic Concepts of Mind Genomics

This chapter gives a detailed overview of the Mind Genomics process, from the implementation point of view. It defines all relevant terms of interest for phases ONE and TWO of the process, and it marks essential differences between the usual approaches to targeted marketing and the Mind Genomics based approach to targeted marketing. Issues are discussed like: How to prepare the polling campaign that determines the mind types on the market, and how to do all the steps needed to place new prospects (potential customers) into one of the existing mind types.