bmw11: dealing with the massive data generated by many-core systems dr don grice

Download BMW11: Dealing with the Massive Data Generated by Many-Core Systems Dr Don Grice

If you can't read please download the document

Upload: mabyn

Post on 07-Jan-2016

27 views

Category:

Documents


2 download

DESCRIPTION

BMW11: Dealing with the Massive Data Generated by Many-Core Systems Dr Don Grice. Title: Dealing with the Massive Data Generated by Many Core Systems. Abstract: - PowerPoint PPT Presentation

TRANSCRIPT

  • New Big Data Brings New Opportunities, Requires New AnalyticsUp to 10,000 Times largerUp to 10,000 times fasterTraditional Data Warehouse and Business IntelligenceData ScaleData ScaleDecision FrequencyData in MotionData at RestTelco Promotions100,000 records/sec, 6B/day10 ms/decision270TB for Deep AnalyticsSmart Traffic250K GPS probes/sec630K segments/sec2 ms/decision, 4K vehicles

  • Petascale Analytics, Appliances and EcosystemDeeper InsightsFaster DecisionsSmarter PlanetBig Data is the new resource. The new opportunity is Big Analytics. Every Smarter Planet solution will depend on it.

    Market leadership in the Era of Analytics will be taken by the first player to deliver high volumes of easy-to-use Smarter Planet solutions.

    Ultimate success will require a Petascale Analytics Appliance and a rich ecosystem of data, algorithms and skills.

  • Maximum Insight Requires Combining Deep and Reactive AnalyticsDirectly integrating Reactive and Deep Analytics enables feedback-driven insight optimizationData ScaleData ScaleDecision FrequencyGovernment and Telco industries are leading this trendTraditional Data Warehouse and Business IntelligenceIntegrationIntegrationReactive AnalyticsRealityFastObservationsActionsHistoryDeep AnalyticsDeepPredictionsHypothesesIntegration

  • Watson IBM Research built a computer system that is able to compete with humans at the game of Jeopardy: Human vs. Machine contest. Named Watson, the computer is designed to rival the human mind Answering questions in natural language poses a grand challenge in computer science, and the Jeopardy! clue format is a great way to showcase: Broad range of topics, such as history, literature, politics, popular culture and science Nature of the clues, requires analyzing subtle meaning, irony, riddles and other complexities Based on the science of Question Answering (QA); differs from conventional search Natural Language / Human Interactions Critical for implementing useful business applications such as: Medical diagnosis Customer relationship management Regulatory compliance Help desk supportFeb. 14 / 15 / 16

    2011 IBM Corporation*

    Compute Intensive Workloads(Traditional HPC)

    2011 IBM CorporationIBM Systems and Technology Group

    Fundamental Issues with Large Scale HPCCompute Intensive Workloads Power Efficiency TCO Programmability and Scaleout Frequency is Plateaued More Parallelism is needed Balanced BWs are required for sustained Perf Shared Memory Model vs I/O Accelerator Model Availability and Reliability More Circuitry is required Technology Scale makes it worse Design for Availability is required Data Management and Cost of Storing/Moving Data Time Steps & Checkpoints Storage Cost, Energy Cost, BW, Latency Life Cycle Management

    2011 IBM Corporation*

    Amount of Data Generated Growing Much FasterThan BW to Store or Retrieve it Example: 100x improvement in Machine Performance

    Core Frequency has Plateaued 100x Performance -> >100x more cores Memory per core ~ constant? -> >100x more memory

    Checkpoint Data Increase >100x Plus frequency may increase due to reliability changes Time Step Data Increase at least 100x? (with Performance)

    Disk and Tape BWs are basically Plateaued (~100MB/s) Compression Methods are not improving much Only provides ~2x BW boost at most in any event Capacity Growing at 20-30% CGR but not BW Amount of Disk/Tape needs to grow >100x to match BW Some relief possible with Write Duty Cycle Utilization Cache locally and take full interval to write it out Pre-stage Reads

    2011 IBM Corporation*

    Example of Data Volume Gap Example of Data Volume Gap Growing for Commercial Users BW Gap is even larger!4% CAGR

    2011 IBM Corporation*

    Data Centric ComputingNetworkDomainRegister StackFunctional UnitsHigh Speed Cluster NetworkLAN/SANWAN?Cluster/SystemMulti-ClusterGrid?SMP BusOS/SMPHigh Speed Cluster NetworkData Set Size Increases DownwardLAN/SAN.. WAN?Disk, Tape?Flash?

    2011 IBM Corporation*

    SUMMARY Data Volume and BW is Exploding in many Areas Multicore/Many Core Compute Intensive Systems Are generating more data and faster than ever before Also using more Memory due to Frequency Stabilization Data Storage BWs are not improving much Balance of Compute to I/O and Storage will need to shift Compute Intensive Workloads will also interact with Data Intensive Workloads in Workflow environments Data Life Cycle Management, Prestaging and Intelligent Writing will become increasingly more important as machines grow in capability

    2011 IBM CorporationIBM Systems and Technology Group

    ...any Questions?Thank you...

    ***Big Data is the new resource. The new opportunity is Big Analytics. Every Smarter Planet solution will depend on it.

    Big Data is really, really big. Up to 10,000 times larger than a typical database. Big Data can also travel very fast, requiring analysis and decisions at a rate up to 10,000 times faster than traditional systems. In addition, with this high volume and high velocity comes a high variety of data, both structured and unstructured, including for example relational tables, documents, log files, sensor readings, blogs, tweets, digital photos and webcam videos.

    To extract business value from these huge volumes of data at rest and in motion, customers need extreme capacity, massively parallel systems running both Reactive Analytics and Deep Analytics. Tightly integrating both analytic styles with bidirectional feedback maximizes actionable insight and makes key processes exceptionally nimble, forming the common analytics architecture for every Smarter Planet solution.

    The movement of Big Data can often swamp the actual processing time for the application, dominating performance cost. This leads to dedicated, long-running Big Analytics systems where the majority of the data resides in the appliance while the applications migrate to it as needed. To handle this extreme workload, we envision a data-centric Peta2 Analytics Appliance capable of applying a Petaflop of computing power over a Petabyte of persistent, byte addressable RAM, likely Phase Change Memory by 2015. By comparison, most data warehouses today weigh in at only a few hundred Terabytes (1 Terabyte = 1000 Gigabytes), and only the worlds largest supercomputers can provide a Petaflop of performance, roughly equivalent to the computing power of 100,000 laptops.

    But even with a Petascale appliance to build with, Smarter Planet solutions will require rich sources of Big Data from the data warehouse, OLTP systems, sensors, and throughout the Internet. New Big Analytics skills will need to be acquired and nurtured, both generic and industry-specific. And to expedite the development and replication of these solutions, a marketplace for Data-as-a-Service and Analytics-as-a-Service will be essential for the complete Big Analytics ecosystem.

    Market leadership in the Era of Analytics will go to the first to profitably deliver high volumes of easy-to-use Smarter Planet solutions. Successful execution will require a Petascale Analytics Appliance supported by a rich ecosystem well populated with data, algorithms and skills. But the key to leadership in Big Analytics is not simply offering an appliance or a set of services. Ultimate success requires the ability to replicate large numbers of solutions at a profit while minimizing time to value, engagement risk, and TCO. The Race to Replication is on.

    *HPC is now being utilized in a wide variety of adjacent markets so there are more potential users than the traditional physicists and chemists. This will require more robust software tools and support. Technology limits are slowing/stopping the increase in single thread performance and causing increased scaleout to generate more application performance. This will also require improved tools and focus on application scaleoutUtility power costs at the data center level is now almost equal to the new equipment spending rates. Getting to Exascale computing levels will also require significant improvements in energy efficiency.pNext addresses the Software issues with higher level abstractions in the languages and portability across machine architecture types by using standards like OpenMP and OpenCL>pNext addresses the power issue by using Heterogeneous/Hybrid architectures where the most power efficient platform for each piece of the application can be used. This hybrid configuration can be at the system level where BlueGene and Cell based systems are used together at the macro level or within a node where different core types are available for specific micro level (instruction or loop level) optimizations.