Hi PEAC Roadmaps 2008 2009 2011 http www

Hi. PEAC Roadmaps 2008 2009 2011 http: //www. hipeac. net/roadmap 2013 2

Impact on society Energy Transport and mobility Education Security Computing Systems Health Aging population Safety Productivity Environment 3

Technology SWOT Europe Market Research objectives Hi. PEAC Recommendations 4

Market trends Market Trends Application Pull • Post-pc devices • Natural interfaces • Interaction with physical world • Data deluge • Intelligent processing • Personalized devices • Sensitive/critical data Business Trends Market Trends • Vertical integration wins • Economy of scale for hardware • Customer lock-in • Customer/user profiling • Fab labs

Post-pc devices

Natural interfaces

Interaction with physical world Infrastructure Core (cloud) Sensory swarm, actuators and real world data Smart house cities, … Mobile access Courtesy Jan M. Rabaey, UC Berkeley, updated for this Hi. PEAC vision 8

Growth of data storage in Exabytes 9

Data Deluge [Paolo Faraboschi, HP] 10

Growth of data storage in Exabytes 11

Intelligent processing More and more applications are not only “number crunching” Recognition, Mining, Synthesis Krste Asanovic et al, “The Landscape of Parallel Computing Research: A View from Berkeley” 12

Personalized services Websites Traffic advice Geolocated Personal video content (multiple viewpoints) Personalized ads Personalized medicine Search filter bubbles!

Sensitive/critical data

Market trends Market Trends Application Pull • Post-pc devices • Natural interfaces • Interaction with physical world • Data deluge • Intelligent processing • Personalized devices • Sensitive/critical data Business Trends Market Trends • Vertical integration wins • Economy of scale for hardware • Customer lock-in • Customer/user profiling • Fab labs

Vertical integration wins Apple Microsoft Google Samsung Amazon … Philips Siemens • Lower integration costs • Complete value chain • Business models for tools are hard to find; big companies can afford to invest in tools • Big companies are more resilient to economic downturns as they have more than one product

Economy of scale for hardware

Customer lock-in • Apple • Microsoft • Google • … • Phone • Phablet • Tablet • Notebook • Desktop • Television • Cloud • Car • …

Customer/user profiling • Facebook • Linked. In • Google • Udacity • Visa • Airlines

Fab labs

Technology SWOT Europe Market Research objectives Hi. PEAC Recommendations 21

Technology trends Technology Trends Constraints • Foundry costs • Power consumption • Complexity Opportunities Technology • • • Third dimension Silicion photonics Wireless connectivity Memory technologies Approximate computing Novel architectures

Foundry costs

Moore’s law: increase in transistor density Data from Kunle Olukotun, Lance Hammond, Herb Sutter, Burton Smith, Chris Batten, and Krste Asanovic 24

The end of Dennard Scaling Parameter (scale factor = a) Classic Scaling Current Scaling Dimensions 1/a Voltage 1/a 1 Current 1/a Capacitance 1/a >1/a Power/Circuit 1/a 2 1/a Power Density 1 a 1/a ~1 Delay/Circuit Source: Krisztián Flautner “From niche to mainstream: can critical systems make the transition? ” 25

Power consumption

Limited frequency increase more cores Data from Kunle Olukotun, Lance Hammond, Herb Sutter, Burton Smith, Chris Batten, and Krste Asanovic 27

Power consumption

Limitation by power density and dissipation 2009: GP CPU = 130 W (45 nm) 2009: Consumer So. C = 10 W 2009: Mobile So. C = 1 W Data from Kunle Olukotun, Lance Hammond, Herb Sutter, Burton Smith, Chris Batten, and Krste Asanovic 30

Energy consumption of ICT Forecast for all ICT domain by the European Commission Servers Television PCs and displays “Business as Usual” 300 TWh in 2010 with 70 TWh for the ~10 million Servers 409 TWh in 2020 with 100 TWh for Servers Note: 1 Nuclear Power Plant ~10 TWh Source: European Commission DG INFSO, Impact of Information and Communication Technologies on Energy Efficiency, final report, 2008 ftp: //ftp. cordis. europa. eu/pub/fp 7/ict/docs/sustainable-growth/ict 4 ee-final-report_en. pdf 31

ICT: 2% of global carbon emissions Source: P. Ranganathan, “Saving the world together, one server at a time…” ACACES 2011 32

Complexity: Achilles heel of software - Parallelism - Scale-out

Technology trends Technology Trends Constraints • Foundry costs • Power consumption • Complexity Opportunities Technology • • • Third dimension Silicion photonics Wireless connectivity Memory technologies Approximate computing Novel architectures

Third dimension

Silicion photonics

Wireless connectivity LTE Advanced = 4 G = 1 Gb/s SATA I = 1, 5 Gb/s SATA II = 6 Gb/s SATA III = 6 Gb/s

Memory technologies

Source: P. Ranganathan, “Saving the world together, one server at a time…” ACACES 2011 40

Approximate computing Probabilistic CMOS Approximate Graphene computing

Novel architectures Declarative programming Neuromorphic computing

Technology SWOT Europe Market Research objectives Hi. PEAC Recommendations 43

SWOT Europe Strengths Weaknesses • • • Europe is full of horizontal specialization • Loss of competitiveness in some domains • Borders and different languages • Lack of venture capitalists • Weak academia-industry link Strong embedded ecosystem Public funding for R&D One of the biggest markets Good education Opportunities • Cost effective customization • Leveraging free/cheap/open infrastructure • Societal challenges • Convergence • Micro- and nano-electronics SWOT Europe 44

Technology SWOT Europe Market Research objectives Hi. PEAC Recommendations 49

Recommendations Strategic Areas ◦ Embedded computing ◦ Data center computing ◦ Mobile computing Cross-cutting challenges ◦ Energy efficiency ◦ System complexity ◦ Dependability Policy recommendation

SA 1: embedded systems Cost effective design of embedded systems Cost-effective certification Secure embedded systems

SA 2: data center computing Low-power micro servers Network and storage I/O and modules

SA 3: mobile computing Supporting immersive/natural interfaces Ensuring security and privacy for personal devices

CC 1: energy efficiency

CC 1: energy efficiency Optimizing data movement and communications Programming heterogeneous parallel processors Developing new computing modalities

CC 2: system complexity

CC 3: dependability

Policy recommendation Supporting virtual verticalization Contingent funding for commercial development Continued funding for academic research International collaboration Developing pilot-line fabrication capabilities Special licenses for collaborative projects

Overview