Robert van der Drift NWO

Transcription

1 Robert van der Drift NWO

2 Call for Proposals: Big Software Software in the Big Data era Matchmaking meeting, 1 July 2015

3 Programme 13:00 Registration 13:30 Introduction Big Software Robert van der Drift, Head of Computer Science (NWO) 13:45 Michiel van Genuchten (COO Vital Health Software) The Impact of Software Jurgen Vinju (Professor Technische Universiteit Eindhoven and group leader Centrum Wiskunde & Informatica) Challenges and Opportunities of Big Software-based Innovation 14:15 Start Pitch & Match sessions 16:00 Drinks & Networking (till 17.00)

4 Introduction Big Software Robert van der Drift Head of Computer Science

5 NWO s role in research funding Ministry of Education, Culture and Science direct government funding 500 M 2,3 Bn universities (incl. medical centres) Ministry of Economic Affairs, Agriculture and Innovation indirect government funding other ministries (e.g. Health, Foreign Affairs, Infrastructure & Environment) NWO institutes private sector and public organisations other knowledge institutes 4

6 The Challenges of Big Software Individual systems have grown to millions of lines of code built from many different technologies. Such systems have come to be known as 'legacy systems' -- systems that resist change. Systems have become more and more inter-dependent, relying strongly on third party components and services, giving rise to systems-ofsystems. The operational context and actual use of software systems has become increasingly complex and unpredictable. In combination, it is increasingly hard to develop software that is reliable, efficient, secure, and evolvable in a timely and cost-effective manner. Big Software welcomes ground-breaking research addressing these challenges.

7 Who can apply Professors, associate professors and assistant professors as well as other senior researchers can apply if they: are employed at a Dutch University or a research institute recognised by NWO, and have at least a master s degree in science or engineering or an equivalent qualification, and have an employment contract for at least the duration of the application procedure and the duration of the research the grant is applied for. Role industrial partner(s): In the project application the industrial and/or public partner(s) must be listed as a co-applicant.

8 Business value for industrial partner(s) Be involved in innovation projects as your part of innovation strategy Gain access to state-of-the-art research and excellent research groups, knowledge and results to be used in product development Meet potential future employees who are the top talents in their research fields

9 Type of project Project team Funded by Contribution Appointed through 2 researchers (PhD-students or postdocs) 1 funded by industrial and/or public parties 1 funded by NWO In cash University 2 researchers (PhD-students or postdocs) 1 funded by NWO 1 fte max 2 employees by industrial and/or public parties In kind University and participating partner(s) 1 PhD-student or postdoc 50% funded by industrial and/or public parties 50% funded by NWO In cash University

10 What can be applied for a) Hiring PhD s and/or 2-year or 3-year postdocs based on fulltime position, (incl. bench fee of 5,000). b) Project-related equipment/software provided the costs are more than 5,000. c) Other project activities, such as knowledge transfer, valorisation, and costs to cover non-scientific personnel, travel/accommodation guest lecturers and organisation of meetings/symposia. Maximum budget for project-related equipment (b) and other project activities (c) is no more than 10% of the total project costs, which will be covered for half of the budget 50%- by NWO. The other half 50% - should be covered by the partner(s) Not eligible for funding: Costs for computers, standard software and other costs which are standard facilities of research institutes General costs for project management and coordination

11 When can be applied The closing date for the submission of application to NWO is Tuesday, 15 September 2015, hours (CET + 01:00). Application consists of: Fact sheet (available at Iris system) Application form in English Letter(s) of Commitment (LoC) a pledge of financial support in cash and in kind contribution

12 Criteria Scientific quality Quality of the consortium Knowledge utilisation More information can be found in Call for Proposals (chapter 4.2). Timeline procedure 15 Sept, hrs Sept-Oct Begin Oct Mid Nov Dec Submission Full Proposal Peer Review Rebuttal applicant Assessment Decision

13 Contact Rosemarie van der Veen-Oei Programme manager Tel. +31 (0) Robert van der Drift Head of Computer Science Tel. +31 (0)

14 More information?

15 Programme 13:00 Registration 13:30 Introduction Big Software Robert van der Drift, Head of Computer Science (NWO) 13:45 Michiel van Genuchten (COO Vital Health Software) The Impact of Software Jurgen Vinju (Professor Technische Universiteit Eindhoven and group leader Centrum Wiskunde & Informatica) Challenges and Opportunities of Big Software-based Innovation 14:15 Start Pitch & Match sessions 16:00 Drinks & Networking (till 17.00)

16 Michiel van Genuchten Straumann Institute

17 09/07/ columns in ASML, Bosch, RealNetworks Philips, Honeywell Hitachi, Uni of Queensland Tomtom, Fujixerox Microsoft, Shell CERN, Oracle, Airbus, JPL, Lint, Bayesian networks, Vodafone India

18 09/07/2015 To date, no significant anomalies have revealed themselves in the flight software.

19 Size of sw in LOC 100M 10M 1M 100K 10K Tokyo ASML railway Oil MR CERN exploration scanner boson Airplane FMS Solaris Mars Cabin swairplane kernell lander Tanzania Bayesian LINT Mobile apps Workflow engine Car Navigation ECU CAR MM player k 1M 100M Volume or unique users in #/year 09/07/2015

20 09/07/2015

21 Compound Annual Growth Rate for sw Software seems to be growing with about 18 % a year Irrespective of application, technology a.s.o. Analysed 50 MLOC closed and 500 MLOC open source No statistical difference in CAGR between the two Relevant for both theory and practice Genuchten, Hatton, IEEE Software, 2012, IEEE Computer 2013 Genuchten, Hatton, Spinellis, to appear, /07/2015

22 sales volume lines of code 09/07/2015 sales volume and software size year sales volume lines of code

23 Suggestions for research To be falsified Defect free rocket science software exists Software grows with about 18 percent a year It s software economics, stupid! Also to be investigated 'legacy-systemen' - ongevoelig voor veranderingen - No Quantifying the benefits of next gen sw technology Explain the 1.18 growth rate (we have some ideas) 09/07/2015

24 Jurgen Vinju Center for Mathematics and Computer Science

25 Software Analysis And Transformation Challenges and Opportunities of Big Software-based Innovation Jurgen J. Vinju Centrum Wiskunde & Informatica TU Eindhoven INRIA Lille Big Software Matchmaking Day July 1st, 2015

26 Go Big Software! [onsoranje.nl] SWAT - SoftWare Analysis And Transformation

27 The Software Medium Printing Press Erasmus Book SWAT - SoftWare Analysis And Transformation

28 The Software Medium Computer Dijkstra Shortest-path SWAT - SoftWare Analysis And Transformation

29 The Software Medium Internet Tim Berners-Lee Web SWAT - SoftWare Analysis And Transformation

30 The Software Medium yesterday s ICT inventions + more and better software = tomorrow s product/services SWAT - SoftWare Analysis And Transformation

31 Software The Innovation Engine from risky products to exploitable services cost-of-development -> cost-of-ownership big bang release -> incremental update from pricy consultants to valuable experts outsourcing -> core business from quantity & complexity to quality & fl exibility constraining people -> supporting people data input -> data discovery SWAT - SoftWare Analysis And Transformation

32 Netherlands = Software Programming Languages Formal Methods Components & Modules Agile Processes Operating Systems Distributed Computing Domain Specifi c Languages Model Driven Engineering Software Architecture Database technology Software Analytics Software Testing The Netherlands: a global leader in software and software engineering SWAT - SoftWare Analysis And Transformation

33 Big Software Big Code Big Process Big Logs Research Better Code Better Process Better Products Complexity => Opportunity SWAT - SoftWare Analysis And Transformation

34 Contextual Software Research [ SWAT - SoftWare Analysis And Transformation

35 Contextual Software Research Great software and software research is contextual, tailor-made Expert, local, domain knowledge is key to success Premature [generalization] is the root of all evil Focus on local urgency and local success factors [Escher] collaborate for the content SWAT - SoftWare Analysis And Transformation

36 Contextual Software Research Building up general SE theory & methods as we go [Jon Sullivan] The goal is incremental, but defi nite, improvement in SE Disruptive innovation is enabled by better software engineering Back to common sense; stop following the hype Use yesterday s and today s assets and experience time-to-market one month sooner? 20% fewer bugs after initial release? how? what if? 50% of the unused features not even developed? developers working on features, not bugs? legacy code an asset instead of a risk? research! SWAT - SoftWare Analysis And Transformation

37 Cross-cutting Contexts Software Contexts are not silo ed in industrial or public sectors Example: High-end Financial Services and Embedded Systems High effi ciency High integration complexity (third-party) High product/service variability Example: Distributed (Big) Data and Meta Programming Systems Intermediate formats Marshalling and transformation Co-evolution of I/O formats and processors SWAT - SoftWare Analysis And Transformation

38 Software for Software Research methods built as (re)usable software automated data collection, analysis, reporting code, process, trace analyses questionnaires & monitors Proof-of-concepts built as software analyzing, transforming, generating, visualizing integrated into existing environments & processes [ Willy Vandersteen] There is no fi eld like ours where knowledge transfer {c,sh,w}ould be organized so directly and faithfully, in either direction only if research has access to the real code, real processes and real logs only if industry has access to full and automated methods and experiments SWAT - SoftWare Analysis And Transformation

39 CWI SWAT Preventing and curing software complexity to enable higher quality software systems, using automated software engineering methods Know-how language engineering source-to-model model-to-source source-to-source mining repositories continuous delivery distributed components Domains embedded systems administrative fi nancial games Connected & collaborative research & education industry & government UvA/HvA/VU/CWI master software engineering TU Eindhoven: Automated Software Analysis SWAT - SoftWare Analysis And Transformation

40 Roadmap ICT Roadmap ICT draft has a fi rst class software theme reliable & fl exible software systems Needs your voiced support Stake our claim that software is a leading factor economically socially academically Contact to enlist support of CIO, CTO, CEO SWAT - SoftWare Analysis And Transformation

41 Yearly Inclusive Excellent speakers Topical posters Save the date Thursday December 3rd Amsterdam Discussion Networking

42 SWAT - SoftWare Analysis And Transformation Big Software a new start for long term collaboration [George Lucas]

43 Andy Zaidman Delft University of Technology

44 Software Engineering Andy Zaidman Big Software Matchmaking Event July 1, 2015

45

46

47

48 People Software Artifacts Running System Software Analytics

49 How to improve reliability, maintainability,? Should we do code reviewing, static analysis,? How should we test? What should we do with our technical debt? Do components with biggest business value change more, show more bugs, How can we speed up continuous deployment?

50 TU Delft coordinating initiative for research, education and training in data science and technology

51 Software for Data Science Cloud Programming: composing computations using mathematically solid foundations Problem: programming multi-core distributed cloud machines with Von Neumann programming languages Problem: data engineers and scientists not trained as software engineers Solution: programming languages that abstract from hardware, close to domain experts Domain-Specific Languages: enabling software engineers to systematically design & apply DSLs reactive extensions interactive extensions Enabling programmability of big data analytics

52

53 Derek Karssenberg Utrecht University

54 PCRaster Research Team Derek Karssenberg Faculty of Geosciences, Utrecht University Geocomputation: simulation of land surface processes Domains: Hydrology (e.g. river flows) Land use change (e.g. bioenergy expansion) Effects of environment on health (e.g. exposure to air pollution) Objectives: Develop concepts and software frameworks Distribute software: PCRaster Team: Software engineers (C++) and PhD students in geoinformatics Domain specialists (water management, health, human geography)

55 Models should be programmable by domain specialists Domain specialists (e.g. hydrologists) are the model builders Need for software providing the building blocks

56 Challenges: (1) modelling heterogeneous systems Problem: Lack of concepts and software frameworks integrating fields and agents

57 Challenges: (2) scalability Big data (e.g. remote sensing) is input to models Requires concurrent execution of models (parallelization, CPU, I/O) Problem: Lack of software framework that allows models built on desktop computers to be run on in a high-performance computing environment (without modification)

58 Our solution Model building framework with built-in support for: Agents and fields Concurrency and parallel I/O, for all hardware platforms

59 Looking for new project partners (companies, research inst.) Our team develops concepts and/or software framework Partner provides problem from a particular domain, e.g. Health & environment (possibility to join Global Geo Health Data Centre in Utrecht) Water management Ecology Crop growth, bioenergy Sensor networks Contact: Derek Karssenberg [email protected]

60 Mark Roest VORtech

61 VORtech: scientific software engineers Established in 1996 Around 25 employees, with academic background in mathematics and IT About 2/3 with a PhD Located in Delft

62 VORtech Services Scientific software engineering Developing scientific software Accelerating and improving scientific software Consultancy on scientific software and mathematics Maintenance of scientific software Specific expertise High Performance Computing Combining sensor observations with models

63 Reason for being here Typical customer code (no relevant proprietary code): 50k to more than 1M lines of code Fortran, C, C++, Pascal/Delphi 1Mb 4Gb data files Interest to learn about techniques for modernization, porting Possible role as intermediary to customers with case studies Mark Roest

64 Yanja Dajsuren Centre for Mathematics and Computer Science

65 Modernizing Big Legacy Software Yanja Dajsuren, CWI NWO Big Software Matchmaking Event Utrecht

66 Different modeling language Different platform More powerful hardware Maintenance cost (engineer KLOC) Legacy software

67 Modernizing legacy software Reo

68 Contact for comments and collaboration: Tel: +31(0) Address: Centrum Wiskunde&Informatica Science Park XG Amsterdam

69 Patricia Lago University of Amsterdam

70 SIMPLE: So*ware Innova1on in complex Eco- systems Research partners Patricia Lago (VU) Paul Grefen & Maryam Razavian (TU/e) Marcel Worring (UvA) Industrial partners (in kind or poten1al) Serge Hollander (OMALA) Sander Klous (KPMG) Maikel Bouricius (GreenIT Amsterdam)?

71 MORE CONTRO PLEASE The Context CONNECTED TRAVELERS : S ELF S ERVICE IN THE ERA OF CONNE TRAVELERS, SELF-SE AND MOBILITY ARE KE UNDERPINNINGS IN T EVOLVING RELATIONS WITH PASSENGERS. IS SUE 2: 2015 CONNECTED TRAVEL: PROXIMITY GATEWAY TO THE INTERNET OF THINGS AS WE STEP TOWARDS THE INTERNET OF THINGS, BEACONS ARE PROVING TO BE A CRUCIAL PART OF THE MIX IN GETTING PROXIMITY AND CONTEXT INFORMATION TO MOBILE DEVICES.

72 The Context A Big- so*ware environment is [K. Dorst]: Open: degrees of visibility and transparency (data, services) Dynamic: Con1nuous change (evolving requirements technologies, opportuni1es) Complex: Shared benefits and shared responsibili1es Networked: Mul1ple stakeholders A field never explored before

73 The Problem How to create so*ware that realizes sustainable innova1on in such a complex environment? Miss opportuni1es (novel business, iden1fy shared op1miza1ons, predict emerging markets) On economic, social, environmental sustainability What is the data (so*ware proper1es, influencing changes, contextual factors like usage) that should be gathered to support sustainable change Miss opportuni1es (iden1fy changing requirements, perform technological adapta1on) On technical sustainability

74 The SIMPLE Approach: ingredients So*ware and Service Engineering 4 sustainability BASE- X: iden1fy innova1on opportuni1e 4 complex eco- systems

75 Look for co- Funding for 2 PhD candidates So*ware Engineering 4 Big- so*ware environments Design Decision Making 4 Sustainable innova1on Visual analy1cs 4 SIMPLE so*ware

76 Contact: Patricia Lago

77 Alexandru Iosup Delft University of Technology

78 Scalable + Available + High Performance Parallel and Distributed Software dr. ir. Alexandru Iosup Parallel and Distributed Systems Group Won IEEE Scale Challenge 2014! 1

79 The Parallel and Distributed Systems group Fun, International, Visible Team also, Award-Winning Join us in 2015! Won IEEE Scale Challenge 2014! 2

80 Scientific Challenges for a Golden Age in ICT How to massivize ICT? Super-scalable, super-flexible, yet efficient ICT infrastructure Data-driven feedback loops for end-to-end automation of large-scale processes Understanding actual use of dynamic, compute- and data-intensive workloads DevOps for evolving, heterogeneous hardware and software under strict performance, cost, energy, reliability, trust, and privacy requirements 3

81 Staff members Big Software for clouds and big data Data-driven feedback loops for scalable, high-performance, efficient operation Meaningful operational logs, including performance and reliability data Open-source software stacks for cloud computing and big data processing, including Hadoop / Spark, Giraph / other graph-processing systems Continuous (re-)deployment of systems of systems (deep stacks) Analysis and action based of heterogeneous datasets and user requirements Analysis of trust and privacy in distributed stacks Benchmarking clouds and big data [email protected] PDS Group, Faculty EEMCS, TU Delft Room HB07.050, Mekelweg 4, 2628CD Delft 4

82 Disclaimer: images used in this presentation obtained via Google Images. Images used in this lecture courtesy to many anonymous contributors to Google Images, and to Google Image Search. Many thanks! 6

83 Tommy van der Vorst Dialogic

84 Research and strategic consultancy Broadband/telecom human capital innovation policy

85 Researchers Customers & stakeholders GIS-data and -tools Dashboards Online reports Analysis modules (Big) data sets Dialogic Platform Crawlers & scrapers Surveys / webforms Text mining Search technology Commercial data sets & feeds Partners

86 Dialogic + Researchers who need a platform that provides userfriendly, real-time and integrated data collection, linkage, analysis and visualisation Software suppliers who can add smart algorithms to the treasure chest, or see new applications of the platform And (obviously): Customers that have a monitoring-, evaluation- or management question, that can be answered through real-time analysis

87 Q & A Tommy van der Vorst MSc Researcher/consultant dialogic.nl/vandervorst nl.linkedin.com/in/tommyvdv [email protected]

88 Aggregate Formatting Formulas lnteraction Metadata Recode Restructure Select Transform Values Variables http i 'api.dialogicinsight nlldata xml

89 Onderzoek en strategisch advies Breedband/telecom onderwijs/arbeidsmarkt innovatiebeleid

90 Onderzoekers Klanten / stakeholders GIS-data en -tools Dashboards Online rapport Analysemodules (Big) datasets Dialogic Platform Crawlers & scrapers Surveys / webforms Tekstmining Zoektechnologie Commerciële datasets/feeds Partners

91 Dialogic + Onderzoekers die een platform nodig hebben waar gebruiksvriendelijke en real-time dataverzameling, koppeling, verwerking en visualisatie bij elkaar komen Software suppliers die slimme algoritmes kunnen toevoegen aan de schatkist, of andere toepassingen zien van het Platform En uiteraard: Opdrachtgevers met een monitorings-, evaluatie- of sturingsvraag hebben die te beantwoorden is met realtime analyse

92 Q & A ir. Tommy van der Vorst Onderzoeker/adviseur dialogic.nl/vandervorst [email protected]

93 Paris Avgeriou University of Groningen

94 7/9/ Big Technical Debt Managing Technical Debt with Big Data Prof. dr.ir. Paris Avgeriou - [email protected] Software Engineering and Architecture Group

95 The problem 7/9/ Technical Debt: Quality Trade-offs Expedient now, expensive later! 50-75% on evolution A necessary evil but must be managed

96 The solution 7/9/ Platform for Managing Technical Debt Source Code Analysis Maching learning Outcome Valuation of internal qualities Accurate effort estimation Actionable figures in dashboard For companies trying to lower maintenance cost

97 Joeri van Leeuwen ASTRON

98 Joeri van Leeuwen A Global Software Telescope for Radio Astronomy Van Leeuwen A Global Software Telescope for Radio Astronomy NWO Big Software

99

100 Fact sheet Further intensification in software, HPC & storage ( Peta -> Exa ) One of the most demanding domains on IT SW/HW require new solutions for scalability, awareness, performance/w, etc. Algorithms Development driven for parallelization / new classes of HPC platforms Relevant for other big data domains: MRI, seismic imaging, remote sensing, &c. For 1B Eur SKA telescope; need to deliver a next-gen processing platform Builds on LOFAR/Westerbork + many opportunities for industry. Working on PPS with several industrial relationships mainly in the software domain so that they qualify for the procurement around In the heart of the ICT Roadmap of Topsectoren, extreme streaming data Business value of PPS is the applicability of the approach to the broad area of big data, streaming data, etc. a very good candidate for further valorization Van Leeuwen Searching for pulsars with LOFAR and fast transients with Apertif NAC Winter Meeting Jan 2015

101 Contacts Dr. Joeri van Leeuwen Astronomer, Principal Investigator Dr. Gert Kruithof Head of R&D Van Leeuwen Searching for pulsars with LOFAR and fast transients with Apertif NAC Winter Meeting Jan 2015

102 Jeroen Keiren Open University of the Netherlands

103 Pitch Big Energy Data Matchmaking event NWO 1 Juli 2015 Christoph Bockisch, Jeroen Keiren, Rody Kersten, Bernard van Gastel, Marko van Eekelen; Open Universiteit, The Netherlands/RU Nijmegen The world-wide total energy consumption is steadily increasing, and energy is becoming a scarce resource. This also affects IT systems, to which a growing percentage of this energy drain is attributed. Furthermore, because hardware costs are decreasing, the operating costs of IT solutions are increasingly determined by energy costs. Reducing the energy consumption of IT systems can therefore offer both an immediate and long-term benefit to both the environment, through a lower consumption of scarce resources, and consumers, by lowering their energy bill. In current practice effort is spent optimizing energy efficiency of hardware. However, the effects of software attract less attention. Still, software plays an important role in the energy consumption of software controlled systems. Imagine I own an energy-efficient car. The actual energy consumption of the car depends on how heavy my right foot is, if I have a heavy foot, this results in a higher fuel consumption. Compare this to a software controlled system: the hardware (the car) can be energy efficient, but the actual consumption depends on the way it is used by the software (the right foot). Our key question is how to make this control software into a responsible driver. We propose to investigate energy consumption of software controlled systems from different perspectives: (1) measure energy consumption of relevant devices at a high frequency generating a large amount of data; (2) use machine learning techniques to derive a model of the energy consumption of the hardware; and (3) use these energy models to determine and optimize the energy consumption caused by the software controlling these systems. There are some existing approaches that consider the separate parts. The research challenge here is to provide an integrated, end-to-end approach and to validate the effectiveness of this process in practical case studies. Currently, the software improvement group (SIG) has offered their support. IBM has shown an interest, and talks are currently ongoing. If you are interested in reducing energy consumption of software controlled systems, contact us! Dr.ir. Jeroen J.A. Keiren [email protected] Prof. dr. Marko van Eekelen [email protected]

104 Pitch Big Energy Data Matchmaking event NWO 1 Juli 2015 Christoph Bockisch, Jeroen Keiren, Rody Kersten, Bernard van Gastel, Marko van Eekelen; Open Universiteit, The Netherlands/RU Nijmegen The world-wide total energy consumption is steadily increasing, and energy is becoming a scarce resource. This also affects IT systems, to which a growing percentage of this energy drain is attributed. Furthermore, because hardware costs are decreasing, the operating costs of IT solutions are increasingly determined by energy costs. Reducing the energy consumption of IT systems can therefore offer both an immediate and long-term benefit to both the environment, through a lower consumption of scarce resources, and consumers, by lowering their energy bill. In current practice effort is spent optimizing energy efficiency of hardware. However, the effects of software attract less attention. Still, software plays an important role in the energy consumption of software controlled systems. Imagine I own an energy-efficient car. The actual energy consumption of the car depends on how heavy my right foot is, if I have a heavy foot, this results in a higher fuel consumption. Compare this to a software controlled system: the hardware (the car) can be energy efficient, but the actual consumption depends on the way it is used by the software (the right foot). Our key question is how to make this control software into a responsible driver. We propose to investigate energy consumption of software controlled systems from different perspectives: (1) measure energy consumption of relevant devices at a high frequency generating a large amount of data; (2) use machine learning techniques to derive a model of the energy consumption of the hardware; and (3) use these energy models to determine and optimize the energy consumption caused by the software controlling these systems. There are some existing approaches that consider the separate parts. The research challenge here is to provide an integrated, end-to-end approach and to validate the effectiveness of this process in practical case studies. Currently, the software improvement group (SIG) has offered their support. IBM has shown an interest, and talks are currently ongoing. If you are interested in reducing energy consumption of software controlled systems, contact us! Dr.ir. Jeroen J.A. Keiren [email protected] Prof. dr. Marko van Eekelen [email protected]

105 Susan Branchett Netherlands esience Center

106 Core escience technologies

107 at t he interface of research and ICT to in1plement escience project s and technology su-table for a braad range of users

108 Business with research question? Susan Branchett Director Business Development

109 Joep de Ligt Hubrecht Institute

110 Developing & sharing bioinformatic pipelines for big genomics Joep de Ligt PhD Dept. Genome Biology Hubrecht Institute

111 A true big data challenge U.S. to analyze DNA from people ($215 million) UK to sequence patients ($160 million) Sequencing is the easy part, analysis is a big challenge

112 Arvados as a show-case Cha llenge: Analyse 8 human genomes within 2 days without buying hardware or writing custom code come: Out 8 genomes in parallel in 1 day and 18 hours in the cloud, after 1 week of setup time

113 Rewriting heritage code

114 Key points Fully Open Source Docker images for software deployment Parallel compute & no IO bottleneck Reproducibility and sharing More on the future of software development: Prins & de Ligt et. al. 7th July Nature Biotechnology

115 Adriënne Mendrik University Medical Center Utrecht

116 Adriënne Mendrik Post-doctoral researcher Image Sciences Institute, UMC Utrecht I m a computer scientist specialized in medical image analysis with 10 years of experience in the field. My current interests lie in finding approaches that bridge the gap between medical image analysis research and clinical practice [email protected]

117 Why is it necessary to bridge the gap? For example: Brain tissue segmentation/quantification in MRI is one of the oldest medical image analysis tasks. Many algorithms have been proposed since 1985 Clinical practice today still qualititative assessment, clinician looks at MRI scan. Medical image analysis is challenging: Requiring highly reliable results While dealing with large variations in patient anatomy/pathology, scanner hardware/software and acquisition protocols.

118 Complex problems have a higher chance of getting solved by joining forces, both in data sharing and combining algorithms. I m currently setting up a collaborative evaluation platform for medical image analysis, that brings together clinical researchers, technical researchers and companies. Collaborators for setting up the platform: Stephen Aylward (Senior Director of Operations, Kitware, North Carolina, USA), Bram van Ginneken (Professor of functional image analysis, Radboud University Nijmegen Medical Centre), and Guido Gerig (Professor of computer science, University of Utah / New York University, USA) Collaborate or brainstorm? Software design is an important consideration for the platform. I m always interested in a good brainstorm about this. Who knows what great ideas might come out. Feel free to contact me: [email protected]