Moving HPC into the Mainstream Addressing the challenges of Mainstream HPC
WHO’S ON THE PANEL Jon Collins / The Register Nick Curcuru / MasterCard Onur Celebioglu / Dell EMC Bryan Betts / Freeform Dynamics
AGENDA HPC coming into the mainstream • HPC challenges – security and beyond • Making HPC part of the solution • Starting out with HPC •
THE DEFINITION OF HPC IS EX EXPANDING AND ACCEL ELER ERATI TING ON PREMISE HYBRID CLOUD PUBLIC CLOUD High Performance Data Traditional HPC Analytics Computationally-intensive Analyzing big data for rapid modeling and simulation insights, real-time results, • Computer-aided design predictive analytics (CAD/CAM/CAE) • Personalized medicine • Weather forecasting • Fraud detection • Oil exploration • Machine Learning Data-centric HPC Genomics • Financial modeling • Signal processing •
MASTERCARD – MOVING THE ANALYSIS CLOSER TO THE TRANSACTION Business need Solution Benefits Turning 2.2B global cards MasterCard worked with Dell • Understand how, when and where customers buy leading 160M transactions/hour, and Cloudera to build a secure, to predictive analytics 52B transactions / year PCI-certified Hadoop cluster. • Increased security and fraud protection via analyzing into intelligence with 1.9M MasterCard's Enterprise Data customer purchasing patterns, affinities and rhythms using machine learning rules applied to those Hub fully conforms to the PCI- • Expand business with anonymized data re: share of wallet compared to transactions. DSS V 2.0 security standards so competition, average ticket and frequency to identify marketing opportunities and it can host PCI datasets and measure return on investment integrate with other systems. “Data privacy and protection is a top priority for MasterCard. As we maximize the most advanced technologies from partners and vendors, they must meet the rigorous security standards we’ve set.” Gary VonderHaar, Chief Technology Officer, Architecture, MasterCard
HPC BECOMING PART OF MAINSTREAM IT EXPANDING THE BOUNDARIES OF WHAT’S POSSIBLE Research Quickly adopting systems for complex scientific analysis Accelerating time-to-insight for drug design, cancer research, agriculture, forensics, Life sciences genomics and bioinformatics Manufacturing Optimizing the performance of software with systems specifically designed for CAE High-performance Exploring the possibilities of machine learning, deep learning and AI with benchmarked data analytics and optimized HDPA configurations. Fueling the algorithms that revolutionize oil and gas research by precisely pinpointing Oil and gas elusive oil and gas reserves Helping customers process trades and transactions faster, determine risks and Financial services opportunities
LOTS OF HPC ACTIVITY – BUT ONLY A MINORITY ARE FULLY HPC-READY Where relevant, how ready are your existing IT systems to handle Do you have HPC/supercomputing in production evolving compute-intensive yet, or are you moving in that direction? requirements over the next 3 years? In production Currently implementing Investigating No activity Unsure Fully Partially Not very Nothing in place
HPC CHALLENGES AND SOLUTIONS – SECURITY, GOVERNANCE AND BEYOND • NYSE offers best and worst examples • Most transactions now machine based • Avalanche effects • Data types and risk factors • Customer-facing information • Governance and compliance • Heal thyself – HPC’s role in security and governance • Anomaly detection • Early warning
DELL EMC TAKES A UNIQUE APPROACH TO HPC Advancing HPC – from • Dell EMC is only 1 the workgroup to TOP500 company in the world with an end- Democratizing HPC – 2 to-end HPC enabling small and portfolio medium businesses Optimizing HPC – 3 • Three focus areas: ensuring balanced price/performance
MODULAR APPROACH – CONCEPTUAL ANALYTICS COMPONENTS USE USE CASE 1 USE CASE 2 USE CASE 3 OPEN SOURCE APPLICATION AND APPLICATION AND APPLICATION AND PROCESS INTEGRATION PROCESS INTEGRATION PROCESS INTEGRATION DATA ANALYTICS DATA ANALYTICS DATA ANALYTICS FOCUS ON FLEXIBILITY; SUBJECT TO CHANGE FROM USE CASE TO USE CASE PREFERENCE DATA STORE DATA STORE DATA STORE … DATA INGESTION/ DATA INGESTION/ DATA INGESTION/ ENGINEERING ENGINEERING ENGINEERING BUSINESS RULES DATA MANAGEMENT AND GOVERNANCE TECHNICAL ENFORCEMENT FOCUS ON STANDARDS; STABLE ACROSS USE CASES HADOOP ENVIRONMENT TECHNOLOGY FOUNDATION COMMERCIAL HARDWARE/INFRASTRUCTURE PRODUCT
STARTING OUT WITH HPC – THINK BIG, START SMALL, SCALE FAST • Understand workloads, business processes and use cases – existing and possible • Build business case based on NPV, not ROI • Create picture of available data – internal and external, collection methods • Culture change– new solutions require new skills and attitudes • Data governance – HPC can be part of solution (e.g. ‘toxic’ data combinations) • Management impact and constraints
DELL EMC HPC INNOVATION LAB Work directly with Dell EMC HPC experts to test drive cutting-edge HPC technologies, and test and tune systems prior to purchase Typical HPC Innovation Lab projects While the list of potential projects is virtually limitless, common projects include: • Cluster comparisons • System parameter sweeps • Graphics processing unit (GPU) test comparisons • Efficiency tuning • HPC network testing • HPC storage system optimization
CLOSING THOUGHTS AND FURTHER READING Dell EMC HPC Innovation lab http://reg.cx/2v8f Dell EMC main HPC page www.dell.com/hpc Dell EMC HPC Innovation lab blogs/best practices www.hpcatdell.com
Thank you for watching
Recommend
More recommend