DESIGN TECHNOLOGY FOR THE TRILLION-DEVICE FUTURE

Size: px

Start display at page:

Download "DESIGN TECHNOLOGY FOR THE TRILLION-DEVICE FUTURE"

Christal Townsend
5 years ago
Views:

1 DESIGN TECHNOLOGY FOR THE TRILLION-DEVICE FUTURE Alberto Sangiovanni-Vincentelli The Edgar L. and Harold H. Buttner Chair of EECS, University of California at Berkeley

2 The Emerging IT Scene! The Cloud! Infrastructural core Sensory swarm Mobile access 2 Courtesy: J. Rabaey

1,000 devices per person by 2025 The Immersed Human Real-life interaction

3 Computers and mobiles to disappear! Predictions: 7 trillions devices servicing 7 billion people! 1,000 devices per person by 2025 The Immersed Human Real-life interaction between humans and cyberspace, enabled by enriched input and output devices on and in the body and in the surrounding environment Courtesy: J. Rabaey

Swarm systems that gather, synthesize and apply information will change the way entire industries operate.

well as related systems including energy Water and chemical treatment.

positively influencing related Congestion systems.

4 Swarm systems that gather, synthesize and apply information will change the way entire industries operate. Smart water Apply monitoring and management technologies to help optimize the availability, delivery, use, and quality of water as well as related systems including energy Water and chemical treatment. Chemicals Smart traffic Use real-time traffic prediction and dynamic tolling to reduce congestion and its byproducts while positively influencing related Congestion systems. Energy Smart energy Analyze customer usage and provide customized products and services that help to boost efficiency from the source through the grid to the end user. Energy grid Carbon emissions Carbon emissions Energy Noise pollution Energy sources Public transportation Smart home 4 Courtesy: A. Fisher, IBM

5 Managing Complexity: Coping with Moore s Law Feature size ( nanometers ) 1000nm 100nm 10nm Intel8080 Intel386 Intel486 Pentium PentiumPro PentiumIII IA-64 1nm Bipolar, NMOS C S CMOS? N D P N S

6 How Did we Cope with Complexity? (ASV, Corsi e Ricorsi: The EDA Story, IEEE Solid State Circuits Magazine, 2010) Methodologies (Freedom from Choice) Abstractions 3,500 Tr Tools 125,000,000 Tr Formalization, Rigor, Discipline

7 General principles Verification complexity is managed by: Abstraction: reduce the number of items by aggregating objects and by eliminating unnecessary details with respect to the goal at hand Decomposition: reduce the number of items to consider by breaking the design object into semi-independent parts (divide et impera) Design Complexity is managed by construction (e.g., automatic layout and logic synthesis): Refinement: Start high in the abstraction layers and define a number of refinement steps that go from the initial description to the final implementation Composition: Assemble designs by composing existing parts

8 Cyber-Physical Systems (CPS) Automotive Avionics Buildings Telecommunications Factory automation Transportation (Air traffic control) Power generation and distribution 8

9 The core of the problem today: CyberPhysical Systems (CPS) Computation and networking integrated with physical processes. The technical problem is managing dynamics, time, and concurrency in networked, distributed computational + physical systems. In the year 2054, the entire defense budget will purchase just one aircraft. Norman Augustine

Modeling Cyber-Physical Systems (Lee, ASV: A framework for comparing models of computation, IEEE

CAD, 1998) Platform 1 Physical Actuator Interface Physical Plant 2 Network Platform 2 Physical

Computation 2 Model Equation-based model Computation 3 Physical Physical Plant 12 Interface

10 Modeling Cyber-Physical Systems (Lee, ASV: A framework for comparing models of computation, IEEE Trans. CAD, 1998) Platform 1 Physical Actuator Interface Physical Plant 2 Network Platform 2 Physical Interface Computation 1 Delay 1 Sensor Computation 4 Sensor Physical Interface Platform 3 Delay 2 Computation 2 Model Equation-based model Computation 3 Physical Physical Plant 12 Interface Actuator Different models of computation Abstraction physical modeling Concept of Time C-code System Sensors Networking Actuators Physical system (the plant) Embedded systems (computation) Courtesy: D. Broman

APPROVED FOR PUBLIC RELEASE. DISTRIBUTION UNLIMITED.

behaviors Planning Constrain: N/A Optimize: Cost Cost Optimization SWaP Optimization SWaP Optimization Programmin g Constrain: Performance Optimize: Cost Budgeting Constrain: Cost Optimize: N/A

Mgmt Source Selection Constrain: Performance Optimize: Cost.

11 APPROVED FOR PUBLIC RELEASE. DISTRIBUTION UNLIMITED. Status Quo: There are several areas where change is needed Design process arbitrarily decomposes system and largely ignores complexity undesired and multimode interactions and emergent system behaviors Planning Constrain: N/A Optimize: Cost Cost Optimization SWaP Optimization SWaP Optimization Programmin g Constrain: Performance Optimize: Cost Budgeting Constrain: Cost Optimize: N/A Cost-centric acquisition process provides dis-incentives to incorporation of flexibility and adaptability into system designs System Functional Specification Execution Power Data & Control Thermal Mgmt Source Selection Constrain: Performance Optimize: Cost MILD Systems Engineerin g Constrain: Performance Optimize: SWaP System Layout Subsystem Design Component Design Verificatio n & Validation Constrain: Performance Optimize: N/A Component Testing Verification & Validation Subsystem Testing MIL STD 499A (1969) systems engineering process: as employed today Conventional V&V techniques do not scale to highly complex or adaptable systems (i.e., those with large or infinite numbers of possible states/configurations) SWaP = Size, Weight, and Power Desirable interactions (data, power, forces & torques) Undesirable interactions (thermal, vibrations, EMI) 11

12 Integration Challenges: Plug and Play? Plug and Pray! 12

The Platform Concept (GSRC: DARPA-SIA) ASV, Quo Vadis, SLD?

Texas Instruments OMAP Meet-in-the-Middle Structured methodology that limits the space of exploration, yet achieves good

design re-use at all abstraction levels An intellectual framework for the complete electronic design process!

13 The Platform Concept (GSRC: DARPA-SIA) ASV, Quo Vadis, SLD? Reasoning About the Trends and Challenges of System Level Design, Proc. of the IEEE, Texas Instruments OMAP Meet-in-the-Middle Structured methodology that limits the space of exploration, yet achieves good results in limited time A formal mechanism for identifying the most critical hand-off points in the design chain A method for design re-use at all abstraction levels An intellectual framework for the complete electronic design process! See AUTOSAR, Intel, National Instruments, Cadence, Synopsys, UTC, GM, Magneti Marelli, ELT, Xilinx,... Platform Mapping Platform Design-Space Export Application Space Application Instance Semantic Platform Platform Platform Instance Architectural Space

14 Platform-Based Design Application Space Architectural Space Application Instance Platform Instance 14 Platform Mapping Platform Design-Space Export Platform: library of resources defining an abstraction layer with interfaces that identify legal connections Resources do contain virtual components i.e., placeholders that will be customized in the implementation phase to meet constraints Very important resources are interconnections and communication protocols

Separation of Concerns: Keep the What Separated from the How (see

C-Code Matlab Dymola CPUs Buses Buses Buses Operating Systems

f1 f3 f2 Mapping Performance Analysis Refinement Platform ECU-1

15 Separation of Concerns: Keep the What Separated from the How (see AUTOSAR) IPs Behavior Components Virtual Architectural Components C-Code Matlab Dymola CPUs Buses Buses Buses Operating Systems Development Process Analysis Specification Implementation Behavior f1 f3 f2 Mapping Performance Analysis Refinement Platform ECU-1 ECU-2 ECU-3 Bus Evaluation of Architectural and Partitioning Alternatives

16 Virtual Design and Refinement

17 Design Re-Use Re-use Library Imported IPs

18 Designing a Swarm System: The Problem Space (TerraSwarm, DARPA-SIA) 18

19 The Swarm as a Platform Apps Home security/emergency Energy-efficient home Health monitoring Unpad SWARM-OS A Mediation Layer Resources Sensors/ Input devs Actuators/ Output devs Networks Storage Computing Presenting a uniform API to Apps Developers (similar to trends in the Cloud) [J. Rabaey, VLSI 11]

20 20 SwarmBox Delivery Service (STARnet Arena, Area A1)

Platform-based Design Environment for Synthetic Biological Systems (D.

biological functions and tools by modifying or integrating

using mathematical modeling to direct the construction towards the

BioBricks Building life from the ground up (Jay Keasling, UCB)

21 Platform-based Design Environment for Synthetic Biological Systems (D. Densmore, Boston University) Synthetic Biology: The creation of novel biological functions and tools by modifying or integrating well-characterized biological components into higher-order systems using mathematical modeling to direct the construction towards the desired end product. BioBricks Building life from the ground up (Jay Keasling, UCB) Development of foundational technologies: Tools for hiding information and managing complexity Core components that can be used in 21 combination reliably

22 22 Final Words of Wisdom

23 Essential Ingredients Library of re-usable components Verified, tested, parametrized Interfaces defined and standardized MODELS!!! For verification, testing, analysis Organized hierarchically with clear relationships established Methodology To develop additional components To define architectures for product families To derive models To assemble components To verify Organization Rigor in adding components and verifying compliance for library insertion Standard work to encode and enforce methodology Education and Training

24 The Way Forward Everything is Connected: Society, Electronic and System Industry facing an array of complex problems from design to manufacturing involving complexity, power, reliability, re-configurability, integration. Complexity is growing more rapidly than ever seen Interactions among subsystems increasingly more difficult to predict Pre-existing systems put to work to provide new services Need work at all levels: Methodology, Modeling, Tools, Algorithms Deep collaboration among Governments, industry, and research centers Different Disciplines : Control, Communication, Computer Science, Electrical Engineering, Mechanical Engineering, Civil Engineering, Chemistry, Biology... 24

Center for Hybrid and Embedded Software Systems. Hybrid & Embedded Software Systems

Center for Hybrid and Embedded Software Systems College of Engineering, University of California at Berkeley Presented by: Edward A. Lee, EECS, UC Berkeley Citris Founding Corporate Members Meeting, Feb.