Accelerating the Pace of Discovery Technical Computing at Intel

Accelerating the Pace of Discovery
Technical Computing at Intel

John Hengeveld
Director of High Performance Computing Strategy
Intel Corporation
@IntelXeon

1

The Compute Continuum

Enterprise
Cloud Service Data Technical
Transactional
Delivery Computing
Computing

Desktops Laptops Netbooks Personal Devices Smartphones Smart TVs Intelligent
& Displays Devices

Technical Computing builds on ideas
to develop new products, experiences and science
2 Copyright © 2011, Intel Corporation. All rights reserved.

Today’s Agenda

Accelerating the pace of discovery:
• Technical computing’s contribution to solving the world’s most pressing
challenges

Crossing the Gap:
• Enabling broader access to high performance computing

A look to the horizon of insight:
• Reaching Exascale and expanding HPC


The World’s Need for Scientific Insight
Has Never Been More Acute

Weather/Climate New Forms of Energy Healthcare


Driving Discovery in Addressing
Global Climate Change
The Challenge

• World population: 9.22B by 2075
• Managing weather, climate, water and
environment prediction is critical to the
long term health of our planet

Scientific Progress

• Multi-day weather prediction weather
prediction achieved
• Ocean current simulation enables sea
current insight and study


Creating New Paths to Fuel Our World
The Challenge
• Accurate exploration of
reservoirs
• Oil and gas production from
complex reservoirs
• Reduced cost of hardware and
maintenances?

Scientific Progress

• Found new oil reserves and developing
new drilling techniques to access them
• Modeling Oil Diffusion and
environmental risks
• Breakthroughs in fusion research


Delivering New Insights in Curing Disease
and Forging New Paths to Health
The Challenge

Cost of first AML genome was
1.6Million

Today cost is ~30K – imagine
when cost drops under 1K.
Personalized medicine?

Scientific Progress

• Genome mapping delivered

• Full body medical imaging
enabling visualization to
improved diagnosis


Big Data, HPC, and Cancer

Alex Bayen, Armando Fox, Michael Franklin,
Michael Jordan, Anthony Joseph, Randy Katz,
David Patterson, Ion Stoica, Scott Shenker

UC Berkeley

September, 2011
8

Big Data is …
Massive
• Facebook: 200-400TB/day:
83 million pictures
• Google: > 25 PB/day processed data

Growing
• More devices (cell phones),
More people (3rd world),
Bigger disks (2TB/$100)

Dirty
• Diverse, No Schema, Uncurated,
Inconsistent Syntax and Semantics

9

“Big Data”: Working Definition
When the normal application of current
technology doesn’t enable users to obtain
timely and cost-effective answers of
sufficient quality to data-driven questions MONEY

QUALITY
Challenge: Use machine learning
TIME
Algorithms, HPC/cloud computing
Machines, and crowd-sourced
People to extract value from Big Data
while decreasing the cost of maintaining it

Interesting Big Data for Academic Research?
10

Big Data Opportunity
The Cancer Genome Atlas (TCGA)
• 20 cancer types, 500 tumors each: 5 petabytes
• David Haussler (UCSC) Datacenter online Oct 1
• OK to place Berkeley cluster next to 5 PB cluster
• Novelty: Academic Access yet Important Big Data

11 Slide from David Haussler, UCSC, “Cancer Genomics,” AMP retreat, 5/24/11

TCGA Potential Impact?

“We fully expect that 10 years from now, each cancer
patient is going to want to get a genomic analysis of
their cancer and will expect customized therapy
based on that information.”
Brad Ozenberger
TCGA program director
“Cracking Cancer's Code”
Time Magazine
June 2, 2011

12

Big Data, Genomics, and Cancer
¼ US deaths, 7M/year worldwide

⅓ US women will get cancer

½ US men will get cancer

Cancer: perversion of normal cell
Limitless growth, evolves, spreads

Cancer is a genetic disease
Accidental DNA cell copy flaws +
carcinogen-based mutations lead to cancer
13

5 Steps to Customized Therapy
Sequencing machine that identifies many 300 base pair segments
1 from a cancer tumor

Create full genome sequence of the cancer tumor from many
2 segments (“alignment”)

3 Verify correctness of this sequence

Insights from comparing many other sequences and treatments to
4 tumor genome

Diagnose and suggest of therapeutic targets for cure or non-
5 progression based on tumor genome comparisons and patient records

14

1. Sequencing Machine Costs
Improving Faster Than Moore’s Law
$1,200,000
$1,000,000
$1,000,000

$800,000

$600,000

$400,000

$200,000
$10,000 $1,000
$0

2007 2009 2012-3
2007 wet lab processing problem => 2012 digital processing problem

Looks like sequencing machines not the bottleneck in speed or cost

15

Information Technology Obstacle?

“There is a growing gap between the generation of massively parallel
sequencing output and the ability to process and analyze the resulting
data. New users are left to navigate a bewildering maze of base calling,
alignment, assembly and analysis tools with often incomplete
documentation and no idea how to compare and validate their outputs.
Bridging this gap is essential, or the coveted $1,000 genome will come
with a $20,000 analysis price tag.”
John D McPherson
“Next-generation gap,”
Nature Methods,
October 15, 2009

16

2 & 3. Full sequence and verification

UCSF prototype using
open source SW, Cloud,
Hadoop, Hypertable,
Berkeley tools

>1 year on PC to <1 day in
the cloud

17

4. Compare Sequences
Machine Learning + Data Analytics

Cloud Programming Frameworks and Storage

(AMP Lab strengths)

5. Clinical Diagnosis
Suggest effective therapeutic targets for cure or stabilization
• Often events are relatively rare mutations, and not identifiable by traditional statistical
methods (solutions lie in long tail of rare mutations: tiny needle in huge haystack)

Use Artificial Artificial Intelligence: (Crowd Sourcing)?
• Imitate success of Astronomy’s Galaxy Zoo or Biology’s FoldIt?

18

An Opportunity or Obligation?
Given increasing genomic databases, next breakthroughs in
cancer fight as likely to come from computer scientists as from
biological scientists?

If it is plausible that CS could help millions of cancer patients live
longer and better lives, as moral people, aren’t we obliged to try?

(If need more motivation, huge future industry for customized,
genome based medicine?)

19

Comp Sci and Pasteur’s Quadrant
Research is inspired by:
Consideration of use?
No Yes
Use-inspired
Pure Basic Basic Research
Quest for Yes Research (Pasteur)
Fundamental (Bohr)
Understanding? Attack Big Data by
Helping Fight Cancer?

Pure Applied
No Research
(Edison)

20 From Pasteur’s Quadrant: Basic Science and Technological Innovation, Donald E. Stokes, 1997
Slide from “Engineering Education and the Challenges of the 21st Century,” Charles Vest, 9/22/09

Driving Future Insight
Ecosystem Products and Research
Collaboration Technology Community

Standards Based Solution Optimize Intel® Optimize Computing to
Optimization and Delivery Solutions to Address serve major Scientific
Emerging Workloads and Research Challenges

Intel Xeon, MIC,
Accelerate Time to Supercomputing to
Common Tools &
Insight move forward Science
Compilers

Intel: Full Engagement on Delivery of
Technical Compute Workload Optimization


Transforming The Path to Discovery:
The Growth of High Performance Computing

Annual Server Processor Shipments Supercomputing in 1997
2010
9,500,000
~10 PFLOPS ~$55K/GFLOP
9,000,000

8,500,000
1,000,000
8,000,000

7,500,000

2,000,000
7,000,000

6,500,000

6,000,000

5,500,000

5,000,000
1,500,000

4,500,000

4,000,000

3,500,000

3,000,000

2,500,000

500,000
~1 TFLOP <$50/GFlop

1995 2005 2015
2010
2000
Performance $/GFLOP

Source: IDC


Improving HPC Solutions for
Continued Transformation

Power, Performance, Parallelism: The Three Pillars of HPC


The Future Intel® Xeon® Processor E5
Codenamed Sandy Bridge-EP

Growing Performance
• Up to 8 cores per socket
• Up to 2X FLOPS with Intel® Advanced
Vector Extensions

Efficient I/O
• Integrated PCIe reduces latency and power
• Platform includes integrated 6Gb SAS for
high performance local memory

Powerful. Advanced Security
Intelligent. • Support for the latest Intel security
features like Intel® Trusted Execution
Technology and Intel® AES New instructions

The Foundation of the Next Generation Datacenter


Intel® Xeon® E5:
Broadest Xeon Product Line

~2X ~20X

Xeon 5500 Xeon E5 Xeon 5500 Xeon E5

Expect to Launch Almost 2X the Cloud and Supercomputing Drive
Designs of Xeon 5500/5600 Unprecedented Initial Demand


Introducing “Koomey’s Law”
2008+2009 laptops
1.E+16
1.E+15 SiCortex SC5832
Dell Dimension 2400
1.E+14
Gateway P3, 733 MHz
1.E+13
Dell Optiplex GXI
IBM PS/2 E + Sun SS1000
1.E+12
486/25 and 486/33 Desktops
1.E+11
Computations per kWh

Macintosh 128k Compaq Deskpro 386/20e
1.E+10 IBM PC IBM PC-AT
Cray 1 supercomputer IBM PC-XT
1.E+09 Apple IIe
DEC PDP-11/20 Altair
1.E+08
SDS 920 8800 Commodore 64
1.E+07
1.E+06 Univac III (transistors)
1.E+05
Univac II
1.E+04
Univac I Regression results:
1.E+03
EDVAC N = 80
1.E+02 Eniac Adjusted R-squared = 0.983
Comps/kWh = exp(0.4401939 x Year - 849.1617)
1.E+01
Average doubling time (1946 to 2009) = 1.57 years
1.E+00
1940 1950 1960 1970 1980 1990 2000 2010


2S Server Power Efficiency
Compelling Energy Efficient Performance Gains

Six
Core
25x
Relative Perf / Watt

32nm
Quad
Core
20x 45nm
Quad
Core
15x 45nm
Quad
Core
10x Dual 65nm
Core
65nm
5x

2004 2005 2006 2007 2008 2009 2010
Xeon® Xeon®
Intel Single Core Intel Multi Core 5500 5600

Outstanding Performance Per Watt Improvement
Source: Intel internal measurements using SPECjbb2005* as of Feb 2009.
Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate performance of Intel products as
measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance. Buyers should consult other sources
of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the
performance of Intel products, visit Intel Performance Benchmark Limitations


Intel’s Many Core
and Multi-core Engines

Intel Xeon:
– Foundation of HPC Performance
– Suited for full scope of workloads
– Industry leading
Multi-core Intel® Xeon® processor at 2.26-3.5 GHz
performance/watt for serial &
photo highly parallel workloads.

MIC Architecture:
– Optimized for highly parallelized
compute intensive workloads
– Common software tools with
Xeon enabling efficient app
readiness and performance
Many Integrated Cores at 1-1.2 GHz tuning
Die Size not to scale


Process Technology Leadership:
The Foundation for Performance Power, and
Parallelism Innovation

22nm
2003 2005 2007 2009 2011

90nm 65nm 45nm 32nm 22nm
A Revolutionary Leap
in Process Technology
SiGe SiGe

Invented 2nd Gen. Invented 2nd Gen. First to
37% Performance
SiGe SiGe Gate-Last Gate-Last Implement Gain at Low
Strained Strained High-k High-k Tri-Gate Voltage*
Silicon Silicon Metal Gate Metal Gate
Strained Silicon >50% Active Power
Reduction at
High-k Metal Gate Constant
Performance*
Tri-Gate

The Foundation for All Computing
Source: Intel
*Compared to Intel 32nm Technology


Today’s Agenda

• High performance computing’s contribution to solving the world’s most
pressing challenges

Crossing the Gap:



Is it Enough?

Lack of Access Remains Critical


Opportunity in Industry:
Benefits of Digital Testing & Refining of Product Designs

Simulation Digital Content Creation

Digital Development Opportunity The Cost of Dedicated Digital
is Acute Development Remains a Barrier
• Lower R&D costs1 • Intel® Xeon® dual-processor workstations
– ~½ the number of physical prototypes standard for professional creators
– ~½ prototyping costs • But access to digital render farms still limited
• Accelerated product delivery2 • External cloud can provide an alternative to
– Competitive advantage dedicated systems for peak load
– Dreamworks3 used 11m hours of external cloud
rendering creating Kung Fu Panda (20% of total)

Full Time Investment for Intermittent Use
Source: 1. Aberdeen group
Source: 2. Aberdeen 2010 Manufacturing Survey
Source: 3. Dreamworks


Bridging the Great Divide: Seamless Access to
Efficient Compute Services on Demand

Our Challenges to Getting There:
• Technical & Social Barriers
• New economic Models for Shared High
Performance Computing
• New Standards


International and Energy and
Nuclear Security Domestic Security Environmental Security

Basic Science Engineering Computing

Lawrence Livermore National Laboratory 34

Basic Science

Cyber Security
NIF Climate

Nuclear
Counterterrorism
Energy Security
Predictive Integrated Codes

Physics and Experimental Verification
Computers
Engineering Models and Validations


Building Energy
Efficiency

Carbon Capture

Addressing barriers of HPC adoption due to Electric Grid
• Lack of expertise
Building new programs through
• Lack of appropriate software application of HPC and
• Cost computational science


International Center for High
Energy Density and Inertial Fusion
Energy Science
National Ignition Facility

 Campus-like environment with
collaborative space
 Building-level security
 Ready access for all partners,
including foreign nationals
 Wireless capability and High Performance Computing
unclassified computing High Performance Computing
Capabilities and Facilities
 Synergy with community plans
for economic growth

Transportation Energy Center
Combustion Research Facility


…engaging with U.S. industry to enhance American economic competitiveness by
promoting the adoption of high performance computing.

 Lowering the barriers for adoption
of HPC:
 High cost of entry
 Lack of appropriate software
HPC
 Shortage of skilled personnel at Labs
100,000+ CPUs

 Deliver true business solutions computer and
for our industrial partners computational
scientists
1,000-10,000 CPUs

 Build and nurture an HPC
innovation ecosystem in the community of industrial
Livermore Valley Open Campus computational users
10-100 CPUs

Lawrence Livermore National Laboratory

Frameworks

Domain Specific Software Ensembles of
calculations

e.g. Uncertainty
Quantification (UQ)
Optimized HPC Libraries Pipeline

Rapid prototyping,
HPC-IC partners development, and
Leading edge deployment
development
Vendors / co-design of reusable Portable and
software scalable software

LLNL programs Competitive
advantage

Lawrence Livermore National Laboratory

Today’s Agenda

• High performance computing’s contribution to solving the world’s most
pressing challenges

Crossing the Gap:



An Insatiable Need For Computing
1 ZFlops

100 EFlops

10 EFlops

1 EFlops

100 PFlops

10 PFlops

1 PFlops Weather
Prediction
100 TFlops

10 TFlops

1 TFlops Genomics
Research
100 GFlops

10 GFlops

1 GFlops
Medical
Imaging
100 MFlops Forecast
1993 1999 2005 2011 2017 2023 2029

Exascale Problems Cannot Be Solved Using the
Computing Power Available Today
Source: www.top500.org


Intel Labs & HPC
Strong Research World Class Research
Partnerships in HPC

Memory
Silicon
Industry Stacking &
Photonics
Technologies

Programm-
Government Security
ability

Interconnect Power
Universities
Technologies Reduction

Delivering Breakthrough
Technologies to Fuel Innovation
* Other names, logos and brands may be claimed as the property of others.


Driving Future Insight
Ecosystem Products and Research
Collaboration Technology Community

Standards Based Solution Optimize Intel® Identify Computing’s
Optimization and Delivery Solutions to Address Role in Scientific and
Emerging Workloads Research Challenges

Intel Xeon, MIC,
Accelerate Time to Top 500 Exascale
Common Tools &
Insight Insight
Compilers

Intel: Full Engagement on Delivery of
Technical Compute Workload Optimization


In Summary

The world’s most pressing challenges continue to be
solved with the help of Intel based technical
computing

The path ahead offers new opportunities in both
continued performance transformation and creating
gateways for new access

Intel is committed to help deliver solutions to meet
these new opportunities

Intel is collaborating across the world to help achieve a
new era of insight .. transforming scientific discovery
once again


Legal Disclaimers
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as
SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those
factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated
purchases, including the performance of that product when combined with other products.
Intel does not control or audit the design or implementation of third party benchmarks or Web sites referenced in this document. Intel encourages all of its
customers to visit the referenced Web sites or others where similar performance benchmarks are reported and confirm whether the referenced benchmarks
are accurate and reflect performance of systems available for purchase.
Relative performance is calculated by assigning a baseline value of 1.0 to one benchmark result, and then dividing the actual benchmark result for the
baseline platform into each of the specific benchmark results of each of the other platforms, and assigning them a relative performance number that
correlates with the performance improvements reported.
SPEC, SPECint, SPECfp, SPECrate. SPECpower, SPECjAppServer, SPECjbb, SPECjvm, SPECWeb, SPECompM, SPECompL, SPEC MPI, SPECjEnterprise* are
trademarks of the Standard Performance Evaluation Corporation. See http://www.spec.org for more information. TPC-C, TPC-H, TPC-E are trademarks of
the Transaction Processing Council. See http://www.tpc.org for more information.
Intel® Virtualization Technology requires a computer system with an enabled Intel® processor, BIOS, virtual machine monitor (VMM) and, for some uses,
certain platform software enabled for it. Functionality, performance or other benefits will vary depending on hardware and software configurations and may
require a BIOS update. Software applications may not be compatible with all operating systems. Please check with your application vendor.
Hyper-Threading Technology requires a computer system with a processor supporting HT Technology and an HT Technology-enabled chipset, BIOS and
operating system. Performance will vary depending on the specific hardware and software you use. For more information including details on which
processors support HT Technology, see here
Intel® Turbo Boost Technology requires a Platform with a processor with Intel Turbo Boost Technology capability. Intel Turbo Boost Technology
performance varies depending on hardware, software and overall system configuration. Check with your platform manufacturer on whether your system
delivers Intel Turbo Boost Technology. For more information, see http://www.intel.com/technology/turboboost
No computer system can provide absolute security under all conditions. Intel® Trusted Execution Technology (Intel® TXT) requires a computer system
with Intel® Virtualization Technology, an Intel TXT-enabled processor, chipset, BIOS, Authenticated Code Modules and an Intel TXT-compatible measured
launched environment (MLE). Intel TXT also requires the system to contain a TPM v1.s. For more information, visit
http://www.intel.com/technology/security. In addition, Intel TXT requires that the original equipment manufacturer provides TPM functionality, which
requires a TPM-supported BIOS. TPM functionality must be initialized and may not be available in all countries.
Intel® AES-NI requires a computer system with an AES-NI enabled processor, as well as non-Intel software to execute the instructions in the correct
sequence. AES-NI is available on Intel® Core™ i5-600 Desktop Processor Series, Intel® Core™ i7-600 Mobile Processor Series, and Intel® Core™ i5-500
Mobile Processor Series. For availability, consult your reseller or system manufacturer. For more information, see http://software.intel.com/en-
us/articles/intel-advanced-encryption-standard-instructions-aes-ni/
Intel processor numbers are not a measure of performance. Processor numbers differentiate features within each processor series, not across different
processor sequences. See http://www.intel.com/products/processor_number for details. Intel products are not intended for use in medical, life saving, life
sustaining, critical control or safety systems, or in nuclear facility applications. All dates and products specified are for planning purposes only and are
subject to change without notice
Copyright © 2011 Intel Corporation. All rights reserved. Intel, the Intel logo, Xeon and Intel Core are trademarks or registered trademarks of Intel
Corporation or its subsidiaries in the United States and other countries. All dates and products specified are for planning purposes only and are subject to
change without notice


Accelerating the Pace of Discovery Technical Computing at Intel

More Related Content

Similar to Accelerating the Pace of Discovery Technical Computing at Intel (20)

More from Intel IT Center (20)

Recently uploaded (20)

Accelerating the Pace of Discovery Technical Computing at Intel