Kranen HVC Keynote 2011
Transcription
Kranen HVC Keynote 2011
Pioneering the Future of Verification A Spiral of Technological and Business Innovation Kathryn Kranen President & CEO, Jasper Design Automation Haifa Verification Conference – December 6, 2011 Kathryn Kranen’s Bio Electrical engineer, ancient ASIC designer • Early user of gate-level simulation 20+ years in the Electronic Design Automation industry • Vice-chairperson of EDA Consortium board of directors • 2005 recipient of the prestigious “Marie R. Pistilli Women in Electronic Design Automation Achievement Award” • 2009 “EE Times’ Top 10 Women in Microelectronics” Multiple patents filed/pending in formal verification domain Page 2 | © 2011, Jasper Design Automation | Confidential Serial Entrepreneur in Verification Domain Formerly Vice President NA Sales - Quickturn Systems • Pioneered the hardware emulation market Formerly President & CEO - Verisity Design, Inc. • Pioneered constrained-random simulation / testbench automation market Currently President & CEO - Jasper Design Automation • Profitable, private EDA company leading the formal property verification domain Page 3 | © 2011, Jasper Design Automation | Confidential Agenda Perspectives on the EDA industry An attempt to demystify the question: Why do some EDA innovations achieve mainstream adoption, while other worthy technologies fizzle? Ideas on future design/verification innovations Page 4 | © 2011, Jasper Design Automation | Confidential Perspectives on the EDA Industry Rewards: • Huge impact on the world, by enabling all electronic devices • Extremely tight collaboration with semiconductor companies • Wide variety of deep technology challenges • Intelligent global workforce with a strong sense of community Challenges: • Small (<$5B), slow-growing industry • Many complicated pieces to the value chain • Historic EDA business models discourage innovation • Not too popular with venture capitalists Page 5 | © 2011, Jasper Design Automation | Confidential Success Rate of EDA Startups is Low It is estimated that only 1 out of 30 to 40 EDA startups achieves a desirable liquidity event – meaning an IPO or high-value acquisition (i.e. employees make money) How can we predict which ones will succeed? Page 6 | © 2011, Jasper Design Automation | Confidential Successful Execution Requires a Spiral of Technological and Business Innovation Technological innovations must move in lockstep with business innovations for successful market adoption Each “step” of industrial usage generates revenue - and real-world feedback - to fuel the next set of innovations By mastering this execution model, organizations can bring about market revolutions – incrementally! Page 7 | © 2011, Jasper Design Automation | Confidential Technological/Business Innovation Spiral Practical Considerations: Setting the right goals (and adjusting them) Conquering market adoption hurdles Tuning the business model to fit the technology A company’s execution model evolves as it matures. Context for this presentation is early years (pre-profit). Page 8 | © 2011, Jasper Design Automation | Confidential Setting The Right Goals And Continuously Refining Them Page 9 | © 2011, Jasper Design Automation | Confidential Classical Business Plan – EDA Style The Idea . . . . . . . . . . . . . Evolutionary? Complementary? Disruptive? My favorite: first complement, then disrupt. Value Propositions . . . . . How valuable will version 1.0 be? Will anyone be willing to use it?... Pay for it? Market Size. . . . . . . . . . . Potential market is huge, but do you have the skills to penetrate and grow that market? Technology Feasibility . . And what flow integration will be required before a real customer will use the solution? Barriers to Competition. . Which of these are also barriers to YOU? Have you considered adoption barriers? Funding Needs. . . . . . . . You will probably need more than you think. Plan on several iterations of “market learning”. Page 10 | © 2011, Jasper Design Automation | Confidential Be Realistic About Best Case Outcome FEATURE ? . . . . . . . . . . Would this capability be more effective if it were embedded in an existing product? PRODUCT ? . . . . . . . . . . Does it solve a big enough problem to justify a separate buying decision by customers? Are boundaries well-defined? COMPANY ? . . . . . . . . . . Can you potentially generate $50M to $100M from this technology? ENTIRE DOMAIN? . . . . Does the core technology have many product-worthy (adjacent) applications? . Can the domain support multiple companies? Business strategies differ dramatically by target outcome Page 11 | © 2011, Jasper Design Automation | Confidential Example: Early Jasper Formal Property Verifica1on Protocol cer)fica)on End-‐to-‐end packet integrity Asynchronous clocking effects Asser)on-‐based verifica)on Page 12 | © 2011, Jasper Design Automation | Confidential Started Here Jasper Today: Solutions to an Array of System-on-chip Development Challenges Property Synthesis Automated asser)on genera)on Iden)fica)on of coverage holes Inference and synthesis of func)onal proper)es from RTL and simula)on waveforms Verifica1on IP Cer)fica)on of AMBA 4/ACE checkers Popular standard protocols Configurable, illustra)ve, op)mized for formal Architecture Valida1on Executable spec Absence of deadlock Cache coherency Higher Capacity Verify complex 100M gate designs Formal Property Verifica1on Protocol cer)fica)on End-‐to-‐end packet integrity Asynchronous clocking effects Asser)on-‐based verifica)on RTL Development Designer-‐based verifica)on w/o testbench Design trade-‐off analysis X-‐propaga)on detec)on and debug Power management verifica)on Post-‐Silicon Debug Failure signature matching Root cause isola)on Candidate cause elimina)on Valida)on of fixes before re-‐spin Interac1ve Debug Modify/create proper)es on the fly to explore design behavior Page 13 | © 2011, Jasper Design Automation | Confidential Started Here Increased Throughput U)lize mul)ple proof engines on parallel compute resources SoC Integra1on Automated register verifica)on Glitch detec)on Mul)-‐cycle path verifica)on Chip-‐level connec)vity Wider Deployment Proliferate across engineering teams with unique adop)on model Find the Right Place in the Value Chain Aim your product at a “Modular Decoupling Point” in the EDA value chain • Too low: product won’t integrate into customer environment • Too high: customers won’t pay for excess value • Source: Christensen, et. al: "Maximizing Returns from Research” Page 14 | © 2011, Jasper Design Automation | Confidential Value vs. Effort – Pick the Right Strategy The ultimate solution will take some time. What will be your go-to-market strategy? • More likely relevant • Get paid for value • Gain experience The basis of most Business Plans ;) High Value / Low Effort High Value / High Effort Value Nobody wants to be here! Low Value / Low Effort Low Value / High Effort Effort Page 15 | © 2011, Jasper Design Automation | Confidential Bull’s Eye Strategy Aim at bigger problems than current standalone tools can address Exposure to real-world challenges yields methods to overcome them Service methods, once documented, become methodology steps Predictable methodology steps are eventually implemented as tool features Page 16 | © 2011, Jasper Design Automation | Confidential Services Unsolved Problems Methodology Tool Tool Evolution Brings Scalability And bigger unsolved problems! Services Methodology Unsolved Problems Tool Page 17 | © 2011, Jasper Design Automation | Confidential Setting the Price for Early Solutions Considerations: Early adopters debug and work around product issues If you get in the door, you can grow the business later A lower price means more potential customers Top semiconductor “logos” help attract VCs, others Page 18 | © 2011, Jasper Design Automation | Confidential Early On, Hold Out for Maximum Dollars Generates income to (partially) fund operations while you are dedicating your precious resources to that customer Brings accountability by testing your value propositions Prevents “false validation” associated with cheap logos Avoids having to raise the price later (very difficult) You need the customer to have lots of skin in the game • The road to success will be hard, and you don’t want it to be easy for your customer to abandon the effort Then do whatever it takes to make early customers successful Page 19 | © 2011, Jasper Design Automation | Confidential Target the Right Level of Automation EDA innovators often aim at “push-button” automation When full automation falls short, there’s no backup plan • Customers deem the solution too risky for production projects • Solution are relegated to less complex but much less valuable applications (refer to Value/Effort slide) An interactive solution is often a better alternative • Can potentially solve bigger problems sooner (with user’s help) • The user’s participation mitigates risk • Empowered users become fanatics who champion your cause! Usability ≠ GUI … often involves very deep technology Page 20 | © 2011, Jasper Design Automation | Confidential OVERCOMING MARKET ADOPTION HURDLES Page 21 | © 2011, Jasper Design Automation | Confidential Typical Market Adoption Hurdles Resistance to change Flow integration issues Risk of inserting the new solution Concerns over startup’s staying power Poor availability of user resources Try to think from the potential customer’s perspective Page 22 | © 2011, Jasper Design Automation | Confidential Find a Compelling Cause for Change Most engineers have resigned themselves to the box they’re living in. Page 23 | © 2011, Jasper Design Automation | Confidential Organizational vs. Personal ROI Organizational ROI benefits a company/project team • Justifies the purchase price • Value must be visible to the executive holding the budget – Think: “Observability at the outputs” Personal ROI benefits the individual user • Adoption is much easier when ultra-busy potential users actually WANT to use the solution • Top-down mandates are very risky Find a way to address both types of ROI Page 24 | © 2011, Jasper Design Automation | Confidential Identify All Essential Ingredients Understand downstream factors that could prohibit use • Language, models, interfaces, user availability, training Make a plan up front to address them • Parallel process to minimize time-to-market • Partner with others if necessary Even one critical missing piece can render a breakthrough “core technology” useless Page 25 | © 2011, Jasper Design Automation | Confidential Verisity Example (1996-1999) Context: Unprofitable Israeli startup selling a testbench automation tool based on a propriety new language, ‘e’ Hurdles We Overcame: • The language barrier, obviously • Blank Page Syndrome (ramp-up problem) • Single Copy Monster (business model issue) • Lack of object-oriented programmers on RTL teams • Availability of models for standard protocols • Shortage of skilled users Page 26 | © 2011, Jasper Design Automation | Confidential Find the Right Early Subset of the Market Tendency is to strike out developing a general-purpose solution, to address a huge market opportunity • Getting the universal solution right can take too long, delay market learning cycles, burn funds, and increase risk of failure A better approach: • Don’t worry about market size while severely channel-limited • Find a segment you can address with early product • Sharpen your value propositions for that set of customers • Generate revenue to fund the next phase of innovations • Iterate the technological/business spiral with new segments until the universal solution is realized! Page 27 | © 2011, Jasper Design Automation | Confidential Quickturn Example: Early 1990s Context: Business plan called for an ASIC Emulator capable of emulating any arbitrary design. • FPGA place & route issues caused race conditions in designs with more than 2 clocks • End result: our ASIC Emulator couldn’t handle any ASICs! Redirected the team to go after x86 processors (single clock designs at the time) • Closed a ~$5M partnership deal with Intel • Built a multi-box system to handle capacity • Grew the processor emulation segment to $100M in 3 years! • Delivered the universal solution for ASIC Emulation, years later Page 28 | © 2011, Jasper Design Automation | Confidential Fine-tuning the Business Model Sales Channel Synergy A product must “fit” its sales and support team and be scalable to the target market Two kinds of EDA products: High Reach and High Touch High Reach Product High Touch Product Sales channel No special skills, maybe sold over the internet Consultative sale, requiring special skills Support required Little Customer-specific methodology Capability Lower Higher Price Lower Higher Mixed into FAM deal OK Page 30 | © 2011, Jasper Design Automation | Confidential Not a good fit Creating a Repeatable Sales Process Requires lots of experimentation to “crack the code” Sales processes vary dramatically – even among functional verification solutions Example questions: • Are evaluations needed? What evaluation scope? • When should you quote price – early or late? • Top-down or bottom-up sale? • Price agreement before evaluation or after? • Sell through central CAD or through individual projects? • Sell early or late in customer’s project cycle? • What style of salesperson is best? Page 31 | © 2011, Jasper Design Automation | Confidential Monetization of Resources A company has limited resources available to generate as much revenue as possible Only a few ways to increase software revenue: • Increase the license consumption per user (or batch process) • Increase the number of users per project • Increase the number of projects per company • Increase the number of companies • Charge more for the licenses you are selling For each employee, ask: • Which revenue parameter is their work impacting? • Is there something better they could be doing? Page 32 | © 2011, Jasper Design Automation | Confidential !"#$%&'())*#+%',)-.+/0'++)123'&)-,'#456,2.76) ! 96:2;$ Example: Jasper Business Model Innovation ! 961<853-$=->-3?-$@=6.AB$C'"D$1-=$;-2=$ ! C*(D$=-3-.28$=2A-E$&%D$233?28$-01234563$1-=$F?4A6G-=$ ! H=6IA2J8-$&$F634-F?K>-$L?2=A-=4$ ! M24A$-L?5A;$I323F53@$53$!"")$ Great business today • >60% annual revenue growth • 97% renewal rate • >50% average expansion per customer • Profitable and self-sustaining Changed Business Model in 2008 ,-.$ /01234563$ 7-3-.28$ !""#$ !""%$ !""&$ !""'$ !""($ !"")$ !""*$ !"+"$ !"++$ But that wasn’t always the case • Had to learn how to compete with almost-free big vendor tools • Found a way to reliably sell high-value, methodology-intensive solutions – Leveraging our agility to rapidly evolve the software • Key business innovation was “Applications Engineer-based Margin Model” Page 33 | © 2011, Jasper Design Automation | Confidential AE Margin-Based Business Model Priorities 1. Identify Customer’s Top Problems; Set Deployment Goals (Applications and Sites) 2. Assess AE & License Requirements 3. Size Deal Using AE Margin Calculation (# AE-Mos. * Revenue/AE-mo.) 4. $ Deliver as Time-Based Software Licenses Plus Allocated Methodology Support Page 34 | © 2011, Jasper Design Automation | Confidential Resource Investment Fits Market Opportunity High level of R&D investment enables continuous delivery of breakthrough solutions AE activities must generate sufficient revenue to fund the company’s operations R&D AE Sales Marketing G&A Sales, marketing, and administrative headcount is kept to a minimum Page 35 | © 2011, Jasper Design Automation | Confidential Future Design/Verification Opportunities A Few Examples That Require Technological/Business Innovation Spiral Design and Debug Breakthroughs Design Verifiability Advisor • Flags hard-to-verify design characteristics as the RTL/HLM is coded, and suggests alternatives or accommodations “Google-desktop” for RTL/HLM design • Pre–caching and indexing information from simulation and/or static analysis to allow "searchable" on-the-fly scenarios • Challenge: Automatic and compact indexing of “pivot” data Page 37 | © 2011, Jasper Design Automation | Confidential Simulation Breakthroughs “Self-aware” simulation • Profiles tests and eliminates wasted cycles • Reduces the number of simulator licenses, but justifies 2x-3x price through savings on machines, power, and data centers Direct controllability on top of existing simulation runs, for “what-if” analysis and coverage High scale symbolic execution using existing simulation and testbench collateral Page 38 | © 2011, Jasper Design Automation | Confidential System-Level Modeling and Verification System = more HW plus SW plus integration • H/W: more integrated content, greater diversity – analog + RF + logic + uP + DSP + memory • S/W: embedded, OS, drivers, libraries, and applications SW/HW Constraints for Concurrent Development • • Software level: adherence to specified hardware constraints Hardware design: adherence to legacy software constraints SW and HW security verification Power event modeling, measurement and verification in the context of hardware plus software Page 39 | © 2011, Jasper Design Automation | Confidential Thank You