CITRIS Scientific Agenda

1
UC Santa Cruz
Center for Information Technology Research in the
Interest of Society Jim Demmel, Chief
Scientist EECS and Math Depts. www.citris.berkele
y.edu
2
Center For Information Technology Research In The
Interest Of Society

• Major new initiative jointly with UC Berkeley, UC
  Davis, UC Merced, UC Santa Cruz,
  LBNL
• Over 100 faculty from 21 departments
• Many industrial partners
• Significant State and private support
• CITRIS will focus on IT solutions to tough,
  quality-of-life related problems
• 3 other such centers CNSI, CalIT2, QB3

3
Committed Support
Founding Corporate Members of CITRIS

• We have received written pledges to CITRIS of
  over 170 million from individuals and
  corporations committed to the CITRIS long-range
  vision
• 100 million from State for facilities
• Significant Federal funding

4
Outline

• Scientific Agenda
• CITRIS Organization

5
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

6
The CITRIS Model
Societal-Scale Information Systems (SIS)
7
Technology Invention in a Social Context Quality
of Life Impact

• Energy Efficiency
• Disaster Response and Homeland
  Defense
• Education

8
Technology Invention in a Social Context Quality
of Life Impact

• Transportation Planning
• Monitoring Health Care
• Land and Environment

9
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

10
Societal-Scale Systems
Secure, non-stop utility Always connected Diverse
components Adapts to interfaces/users
11
Smart DustMEMS-Scale Sensors/Actuators/Communicat
ors

• Create a dynamic network of power-aware sensors
  for
• Current off-the-shelf design
• 512 bytes RAM, Radio with 10-100 ft range
• TinyOS for programming

Temperature Humidity Pressure Position Acceleratio
n
Light Sound Magnetism Chemicals Biological Agents
12
(No Transcript)
13
Ad-hoc sensor networks work

• 29 Palms Marine Base, March 2001
• 10 Motes dropped from an airplane landed, formed
  a wireless network, detected passing vehicles,
  and radioed information back
• Intel Developers Forum, Aug 2001
• 800 Motes running TinyOS hidden in auditorium
  seats started up and formed a wireless network as
  participants passed them around
• tinyos.millennium.berkeley.edu

14
Smart Dust Goes National

• Selected as DARPA networked embedded system tech
  open platform (NEST)
• Over 5000 Motes used or shipped to other groups
• Academia UCSD, UCLA, USC, MIT, Rutgers,
  Dartmouth, U. Illinois UC, NCSA, U. Virginia, U.
  Washington, Ohio State
• Industry Intel, Crossbow, Bosch, Accenture,
  Mitre, Xerox PARC, Kestrel
• Government Wright Patterson AFB, NCSC
• Ongoing training courses

15
Micro Flying Insect

• Collaboration with Biologist Dickinson

16
Synthetic Insects(Smart Dust with Legs)

• Goal Make silicon walk.

• Articulated Legs
• Size 1-10 mm
• Speed 1mm/s

17
MEMS Technology Roadmap (DARPA)
2010
MEMS Single Molecule Detection Systems
2005
MEMS Rotary Engine Power System
2004
MEMS Micro Sensor Networks(Smart Dust)
2003
MEMS Mechanical Micro Radios
MEMS Immunological Sensors
2002
18
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

19
What is Disaster Response?

• Sensors installed near critical points
• Sensors measure
• Motion (normal deterioration vs serious damage)
• Occupancy (where are people?)
• Fire, heat, chemicals, biological agents
• Sensors report location, kinematics of damage
  during and after an extreme event
• Guide emergency personnel
• Assess structural safety without deconstructing
  building

20
Many Scenarios
21
Seismic Monitoring of Housing

• Many such buildings collapsed or were severely
  damaged in the 1994 Northridge Earthquake.
• Experimental evaluation of a full-scale structure
  on the Richmond Field Station shake table.
• Part of the CUREe-Caltech Tuck-Under Parking
  Apartment Building Experiment

22
Seismic Monitoring of Buildings Before CITRIS
23
Seismic Monitoring of Buildings With CITRIS
Wireless Motes
70 each
24
Mote (ADXL202) vs. Traditional Piezo Accelerometer
Time Domain Comparison
Frequency Domain Comparison
25
Tokachi Port, Hokkaido
Blast-induced Liquefaction Test
26
theirs
ours
theirs
theirs
ours
27
400 Came to Watch
28
Post-Blast Liquefaction
29
A commercial product

• Crossbow CN4000 Wireless Structural Monitoring
  System
• 3D Accelerometer
• 12 bits of resolution, up to 2G
• Temperature
• -40o C to 85o C, to within ?2o C
• Wireless communication
• 1 mile line-of-site range
• www.xbow.com

30
Future Disaster Response Work

• Golden Gate Bridge
• Wind, seismic, security monitoring
• Masada
• King Herods Palace
• Seismic (tourist) monitoring

31
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

32
The Inelasticity of Californias Electrical
Supply
Power-exchange market price for electricity
versus load (California, Summer 2000)
33
How to Address the Inelasticity of the Supply

• Reduce demand, or spread demand over time
• Make cost of energy
• visible to end-user
• function of load curve
• Real-time pricing
• Phase 1 Expose energy usage to user helps
  eliminate waste
• Phase 2 Expose real-time prices to user
• Phase 3 Automatic control to optimize price,
  safety, user comfort, other economic goals
• Improve efficiency of generation and distribution
  network (supply side)

Enabled by Information!
34
Cory Hall Energy Monitoring Network

• 50 nodes on 4th floor
• 30 sec sampling
• 250K samples to database over 6 weeks
• Moved to Intel Lab come play!

35
Control of HVAC systems

• Simulation results assuming multiple sensors
• Hot August day in Sacramento
• Underfloor HVAC saves 46 of energy

36
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

37
Habitat Monitoring on Great Duck Island

• Enable researchers anywhere in the world to
  engage in non-intrusive monitoring of sensitive
  wildlife and habitats
• Study breeding cycle of Leach's Storm Petrel
  
  
  

38
Duck Island System Architecture
39
Duck Island Sample Data

• Light, Temperature, Infrared, Humidity, Power
• Live data at www.greatduckisland.net

40
Monitoring Mogau Caves, China

• Location of ancient cave paintings
• Goal monitor humidity, other factors that could
  damage paintings
• Supported by
• Dunhuang University
• Osaka University, Dept of Global Architecture
• Getty Foundation

41
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• Disaster Response
• Energy Efficiency
• Environmental Monitoring
• Education
• New Application Areas
• High Level Building Blocks

42
Education Goals

• UC Merced curriculum collaboration
• New UC campus to open
• Tele-laboratories and smart classrooms
• Mechanical Rapid-Prototyping, MEMS, Microlab,
  Robotics
• Masters degrees for professionals
• New graduate courses
• Discussions with Chinese Ministry of Education

43
UC Merced (1)

• New campus to open in 2004
• Help accommodate 50 growth in UC
• Goal Export Berkeleys curriculum to Merced
• Start with programming courses
• Not Distance Learning Still need local
  instructors
• Replaces lectures by directed on-line student
  team work
• Instructor monitors teams, gives short directed
  lectures
• Student Portal
• Reading, problem solving, discussion, quizzes
• Course Builder and Customizer for faculty
• Course database to support upgrading and
  customizing

44
UC Merced (2)

• Summer 2002 at Berkeley
• CS3 (Introduction to Symbolic Computing for Non
  Majors)
• Results better exam results, high ratings
• Fall 2002 at Berkeley
• CS3 again, self-paced too
• Spring 2003 at Merced Community College
• Local instructor, Support from Berkeley staff
• 2004
• Main Merced campus to open
• Other courses available

45
(No Transcript)
46
(No Transcript)
47
Masters Degrees for Professionals

• Management of Technology (MOT)
• High performance Communication Networks
• Wireless Systems
• Embedded Computing
• MEMS
• Internet-based Design, Manufacturing, and Commerce

48
Management of Technology (MOT)

• ME221
• High Tech Product Design and Rapid Manufacturing
• Taught in campus TV studio
• Webcast to Intel, Sony and NEC
• 12 off-campus students, many more on campus
• Designed products
• Sent files to Berkeleys CyberCut/CyberBuild
  system
• Custom, Internet-based manufacturing
• See mot.berkeley.edu

49
ME221 Project Examples

• Summit

• !ntro

• Coachs companion

• bentoBox

50
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• New Application Areas
• Electronic Cultural Atlas Initiative (ECAI)
• Berkeley Laboratory in Experimental Economics
  (XLAB)
• High Level Building Blocks

51
Electronic Cultural Atlas Initiative (ECAI)

• Collaborative project which combines global
  mapping, imagery, and texts
• TimeMap to view events, artifacts, images by time
  and place
• Headquartered at Berkeley (Lancaster, Chinese
  Studies)
• A few sample projects (from 300)
• Time Map Korea
• Silk Road Atlas
• South Asian Animation
• Great Britain Historical GIS
• See www.ecai.org

52
Electronic Cultural Atlas Initiative (ECAI)

• Proposed CITRIS Collaboration
• Authoring tools tailored to the needs of specific
  disciplines for course development
• Use of standards for documentation
• Training programs for the use of the tools and
  the development of documentation
• ECAI pilot group focused on Chinese studies
• history
• art history
• language study
• Need for servers

53
Berkeley Laboratory in Experimental Economics
(XLAB)

• Auerbach, Gilbert, Akerlof (Nobel Prize 2001)
• Joint between Economics and Business School
• Experimental Economics becoming major methodology
• Experimentally evaluate economic assumptions,
  theories
• How real people make economic decisions
• Why do some products succeed, others not?
• Need servers

54
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• New Application Areas
• High Level Building Blocks
• Scope
• Project descriptions

55
Societal-Scale Information System - SIS
56
Desirable SIS Features - Problems to solve

• Integrates diverse components seamlessly
• Easy to build new services from existing ones
• Adapts to interfaces/users
• Non-stop, always connected
• Secure

57
Outline Scientific Agenda

• Scientific Agenda Overview
• Low Level Hardware and Software Building Blocks
• Recent progress in large scale applications
• New Application Areas
• High Level Building Blocks
• Scope
• (A few) Project descriptions

58
Projects

• ROC (Recovery Oriented Computing)
• Patterson, Fox (Stanford)
• Oceanstore and Tapestry
• Kubiatowicz, Joseph
• OSQ (Open Source Quality)
• Aiken, Henzinger, Necula
• High Productivity Software
• Yelick, Demmel

59
Recovery Oriented Computing (ROC)

• Dave Patterson and a cast of 1000s
• Aaron Brown, Pete Broadwell, George Candea,Mike
  Chen, James Cutler, Patricia Enriquez, Prof.
  Armando Fox, Emre Kiciman, Matthew
  Merzbacher, David Oppenheimer, Naveen Sastry,
  William Tetzlaff, Jonathan Traupman, and Noah
  Treuhaft
• U.C. Berkeley, Mills College, Stanford
  University
• October 2002

60
Learning from others Bridges

• 1800s 1/4 iron truss railroad bridges failed!
• Safety is now part of Civil Engineering DNA
• Techniques invented since 1800s
• Learn from failures vs. successes
• Redundancy to survive some failures
• Margin of safety 3X-6X vs. calculated load
• To hide errors in building material,
  construction, design, and use
• What is CSE version of safety margin?

61
Margin of Safety in CSE?

• Like Civil Engineering, never make dependable
  systems until add margin of safety (margin of
  ignorance) for what we dont (cant) know?
• Today design to tolerate expected (HW) faults
• RAID 5 Story
• Operator removing good disk vs. bad disk
• Temperature, vibration causing failure before
  predicted on data sheets
• CSE Margin of Safety Tolerate human error in
  design, in construction, and in use?
• Perhaps we need to over engineer to deliver
  what people expect

62
Recovery-Oriented Computing Philosophy

• If a problem has no solution, it may not be a
  problem, but a fact, not to be solved, but to be
  coped with over time
• Shimon Peres (Peress Law)
• People/HW/SW failures are facts, not problems
• Recovery/repair is how we cope with them

• ROC also helps with maintenance/TCO
• since major Sys Admin job is recovery after
  failure

63
MTTR more valuable than MTTF?

• Threshold gt non-linear return on improvement if
  recovery time drops below threshold
• 8 to 11 second abandonment threshold on Internet
• 30 second NSF client/server threshold
• Ebay 4 hour outage, 1st major outage in year
• More people in single event worse for reputation?
• One 4-hour outage/year gt NY Times gt stock?
• 250 people in a single plane is front page news
  1 person per day in a planes is not news, even
  though more die per year in general aviation than
  in commercial
• MTTF normally predicted vs. observed
• Include environmental error operator error, app
  bug?
• Much easier to verify MTTR than MTTF!

64
Five ROC Solid Principles

• Given errors occur, design to recover rapidly
• Extensive sanity checks during operation
• To discover failures quickly (and to help debug)
• Report to operator (and remotely to developers)
• Tools to help operator find, fix problems
• Since operator part of recovery e.g., hot swap
  undo graceful, gradual SW upgrade/degrade
• Any error message in HW or SW can be routinely
  invoked, scripted for regression test
• To test emergency routines during development
• To validate emergency routines in field
• To train operators in field
• Recovery benchmarks to measure progress
• Recreate performance benchmark competition

65
Projects

• ROC (Recovery Oriented Computing)
• Patterson, Fox (Stanford)
• Oceanstore and Tapestry
• Kubiatowicz, Joseph
• OSQ (Open Source Quality)
• Aiken, Henzinger, Necula
• High Productivity Software
• Yelick, Demmel

66
OceanStoreGlobal-Scale Persistent Storage
67
OceanStore Context Ubiquitous Computing

• Computing everywhere
• Desktop, Laptop, Palmtop
• Cars, Cellphones
• Shoes? Clothing? Walls?
• Connectivity everywhere
• Rapid growth of bandwidth in the interior of the
  net
• Broadband to the home and office
• Wireless technologies such as CMDA, Satelite,
  laser

68
Questions about information

• Where is persistent information stored?
• Want Geographic independence for availability,
  durability, and freedom to adapt to circumstances
• How is it protected?
• Want Encryption for privacy, signatures for
  authenticity, and Byzantine commitment for
  integrity
• Can we make it indestructible?
• Want Redundancy with continuous repair and
  redistribution for long-term durability
• Is it hard to manage?
• Want automatic optimization, diagnosis and
  repair
• Who owns the aggregate resouces?
• Want Utility Infrastructure!

69
Utility-based Infrastructure

• Transparent data service provided by
  federationof companies
• Monthly fee paid to one service provider
• Companies buy and sell capacity from each other

70
OceanStore Assumptions

• Untrusted Infrastructure
• The OceanStore is comprised of untrusted
  components
• Only ciphertext within the infrastructure
• Responsible Party
• Some organization (i.e. service provider)
  guarantees that your data is consistent and
  durable
• Not trusted with content of data, merely its
  integrity
• Mostly Well-Connected
• Data producers and consumers are connected to a
  high-bandwidth network most of the time
• Exploit multicast for quicker consistency when
  possible
• Promiscuous Caching
• Data may be cached anywhere, anytime
• Optimistic Concurrency via Conflict Resolution
• Avoid locking in the wide area
• Applications use object-based interface for
  updates

71
First Implementation Java

• Event-driven state-machine model
• Included Components
• Initial floating replica design
• Conflict resolution and Byzantine agreement
• Routing facility (Tapestry)
• Bloom Filter location algorithm
• Plaxton-based locate and route data structures
• Introspective gathering of tacit info and
  adaptation
• Language for introspective handler construction
• Clustering, prefetching, adaptation of network
  routing
• Initial archival facilities
• Interleaved Reed-Solomon codes for fragmentation
• Methods for signing and validating fragments
• Target Applications
• Unix file-system interface under Linux (legacy
  apps)
• Email application, proxy for web caches,
  streaming multimedia applications

72
OceanStore Conclusions

• OceanStore everyones data, one big utility
• Global Utility model for persistent data storage
• OceanStore assumptions
• Untrusted infrastructure with a responsible party
• Mostly connected with conflict resolution
• Continuous on-line optimization
• OceanStore properties
• Provides security, privacy, and integrity
• Provides extreme durability
• Lower maintenance cost through redundancy,
  continuous adaptation, self-diagnosis and repair
• Large scale system has good statistical properties

73
Oceanstore Prototype Running with 5 other sites
worldwide
74
Projects

• ROC (Recovery Oriented Computing)
• Patterson, Fox (Stanford)
• Oceanstore and Tapestry
• Kubiatowicz, Joseph
• OSQ (Open Source Quality)
• Aiken, Henzinger, Necula
• High Productivity Software
• Yelick, Demmel

75
OSQ Open Source Quality

• Goals Automatic analysis of software for
• Finding bugs
• Checking specifications
• Of a at least simple properties
• Help with writing specifications
• Focus
• Large, ubiquitous systems programs
• Linux kernel, sendmail, apache, etc.

76
Tools

• CCured
• Automatically enforce memory safety for C
• Array index out of bounds, wild pointer
  dereferences
• CQual
• Specification and checking of system-specific
  properties
• Locking, file handling, ordering of method calls,
  
• CHIC and MOCHA
• Interface compatibility checking
• Automatic verification of interface protocols
• BLAST
• Software model checker
• E.g., for checking complex control-flow in device
  drivers
• www.cs.berkeley.edu/weimer/osq

77
Ccured - Type-Safe programming in C

• Memory safety
• must have property for reliability and security
  (50 of reported vulnerabilities are due to
  buffer overruns)
• C was designed to be flexible not safe!
• But most C programs use dangerous C features
  benignly
• CCured
• Analyzes the program statically to find benign
  pointer use
• Inserts run-time checks where static analysis
  fails
• Run-time overhead of 50 (unlike 10x for Purify)
• Prototype works for large programs
• Sendmail, bind, openssl, SpiderMonkey engine,
  Apache modules
• Requires some intervention (similar to porting)
• See http//www.cs.berkeley.edu/necula/ccured

78
CQual Extending Standard Types

• Problem Many unchecked properties
• Is the lock acquired?
• Is the file open?
• Idea User-defined type qualifiers

• CQual checks such properties for C programs
• Distribution available
• Found many bugs in device drivers
• Found many security vulnerabilities

const int locked spinlock_t open FILE
const int locked spinlock_t open FILE
79
CHIC and MOCHAInterface Compatibility Checking
Objective Automatic verification of
compatibility between interface protocols of
hardware and software components
Approach -Interface protocols expose more
information than data types The interface
protocol of a file server may specify that the
read-file method cannot be called before the
open-file method has been called. The
interface protocol of a bidirectional bus may
specify that no two clients can write to the
bus at the same time. -Verifying interface
protocol compatibility lies in difficulty between
type checking and full behavioral verification
80
Interface Compatibility Checking
Tools For software interfaces CHIC (extension
of Jbuilder) For hardware interfaces
MOCHA Applications Software interfaces TinyOS
(an OS for adhoc networking) Hardware
interfaces PCI bus and clients Future
Plans -Check conformance of implementation with
interface -Interface protocols with real-time
and resource constraints
81
Projects

• ROC (Recovery Oriented Computing)
• Patterson, Fox (Stanford)
• Oceanstore and Tapestry
• Kubiatowicz, Joseph
• OSQ (Open Source Quality)
• Aiken, Henzinger, Necula
• High Productivity Software
• Yelick, Demmel

82
Tools for High Productivity Computing
Kathy YelickU.C. Berkeley

• http//www.cs.berkeley.edu/yelick/

83
HPC Problems and Approaches

• Parallel machines are too hard to program
• Users left behind with each new major
  generation
• Efficiency is too low
• Even after a large programming effort
• Single digit efficiency numbers are common
• Even on sequential machines
• Approach
• Titanium
• Modern (Java-based) language that provides
  performance transparency
• Kathy Yelick, Susan Graham, Paul Hilfinger, Phil
  Colella
• Bebop Berkeley Benchmarking and Optimization
  group
• Kathy Yelick, Jim Demmel
• Unified Parallel C (UPC)
• Global address space language based on C
• Commercial support (HP, Cray,)

84
Global Address Space Programming

• Intermediate point between message passing and
  shared memory
• Program consists of a collection of processes.
• Fixed at program startup time, like MPI
• Local and shared data, as in shared memory model
• But, shared data is partitioned over local
  processes
• Remote data stays remote on distributed memory
  machines
• Processes communicate by reads/writes to shared
  variables
• Note These are not data-parallel languages
• Heroic compilers not required
• Examples are UPC, Titanium, CAF, Split-C
• http//upc.nersc.gov
• http//titanium.berkeley.edu/

85
Titanium Overview

• Object-oriented language based on Java with
• Scalable parallelism
• SPMD model with global address space
• Multidimensional arrays
• points and index sets as first-class values
• Immutable classes
• user-definable non-reference types for
  performance
• Operator overloading
• by demand from our user community
• Semi-automated memory management
• uses memory regions for high performance

86
Serial Java Performance
87
Serial Java PerformancePentium 4, 1.45 GHz
88
Heart Simulation

• Problem compute blood flow in the heart
• Modeled as elastic structure in incompressible
  fluid
• The immersed boundary method Peskin and
  McQueen.
• 20 years of development in model
• Possible applications in the design of artificial
  heart valves
• Implemented as a general tool for fluid flow with
  elastic structures
• Written in Titanium
• Use Java features for
  extensibility
• Applied to heart, inner ear
• Parallel implementation
  for
  shared/distributed memory

Image from PSC
89
AMR Gas Dynamics

• Adaptive mesh refinement (AMR)
• Places more computation where there is more
  activity
• Uses tree of block-structured meshes
• Gas Dynamics code in AMR
• Developed by McCorquodale and Colella
• 3D supported
• 2D example Mach-10 shock on solid surface
  
  at oblique angle

90
Summary

• Global address space languages offer alternative
  to MPI for large machines
• Easier to use shared data structures
• Recover users left behind on shared memory?
• Performance tuning still possible
• Implementation
• Small compiler effort given lightweight
  communication
• Portable communication layer GASNet
• Difficulty with small message performance on IBM
  SP platform

91
Context High-Performance Libraries

• Application performance dominated by a few
  computational kernels
• Today Kernels hand-tuned by vendor or user
• Performance tuning challenges
• Performance is a complicated function of kernel,
  architecture, compiler, and workload
• Tedious and time-consuming
• Successful automated approaches
• Dense linear algebra PHiPAC, ATLAS
• Signal processing FFTW, SPIRAL, UHFFT

92
Tuning pays off ATLAS
Extends applicability of PHIPAC Incorporated in
Matlab (with rest of LAPACK)
93
Tuning Sparse Matrix Kernels

• Optimizations depend on
• Machine characteristics (as in dense case)
• Nonzero pattern in the sparse matrix
• Performance tuning issues in sparse linear
  algebra
• Indirect, irregular memory references
• High bandwidth requirements, poor instruction mix
• Performance depends on architecture, kernel, and
  matrix
• How to select data structures and implementations
  at run-time
• Typical performance lt 10 machine peak
• Our approach to automatic tuning for each
  kernel,
• Identify and generate a space of implementations
• Search the space to find the fastest one (models,
  experiments)

94
Machine Profiles Computed Offline
Register blocking performance for a dense matrix
in sparse format.
333 MHz Sun Ultra 2i
500 MHz Intel Pentium III
375 MHz IBM Power3
800 MHz Intel Itanium
95
Register Blocked SpMV Performance Ultra 2i
(See upcoming SC02 paper for a detailed
analysis.)
96
Bebop Summary

• Bebop project applying these techniques and other
  optimizations to a number of sparse matrix
  kernels
• Further performance improvements to
  sparse-matrix-vector multiply
• Symmetry (up to 1.5 2x speedups)
• Diagonals, block diagonals, bands (1.2 2x)
• Splitting for variable structure (1.3 1.7x)
• Reordering to create dense structure (1.7 x)
• Cache blocking (1.5 4x)
• Multiple vectors (2 7x)
• And combinations
• How to choose optimizations and tuning parameters
• Sparse triangular solve (1.2 1.8x)
• Higher level Kernels
• yATAx, yAATx (up to 3x)
• Powers (yAkx), sparse triple-product (RART),
  (future work)

97
More Projects

• Millennium and PlanetLab
• Culler, Kubiatowicz, Stoica, Shenker
• www.planet-lab.org/
• www.millennium.berkeley.edu/
• Sahara
• Service Architecture for Heterogeneous Access,
  Resources, and Applications
• Katz, Joseph, Stoica
• sahara.cs.Berkeley.edu/
• Security
• Wagner, Tygar
• Visualization
• Hamann, Joy, Max, Staadt
• CAD for MEMS
• Demmel, Govindjee, Agogino, Pister, Bai
• www-bsac.EECS.Berkeley.EDU/cadtools/sugar/sugar/

98
Outline

• Scientific Agenda
• CITRIS Organization

99
(No Transcript)
100
Current and Near Term Space

• Intel Lab in Power Bar Building on Shattuck
• Hearst Mining (Early 2003)
• BID (Berkeley Institute of Design)

101
The New CITRIS Building

• Construction will begin in summer 2003
• Architectural plans are well underway
• It will house the Microfabrication Laboratory
• Remaining space will be allocated to other CITRIS
  related projects
• Including Corporate Visitors

102
(No Transcript)
103
CITRIS-Affiliated Research Activities

• Berkeley Sensor and Actuator Center (BSAC) (14
  faculty, 100 students)
• Designs sensors and actuators
• Microfabrication Laboratory (71 faculty, 254
  students)
• Fabricates chips
• Berkeley Wireless Research Center (BWRC) (16
  faculty, 114 students)
• Designs low-power wireless devices.
• International Computer Science Institute (ICSI)
  (5 faculty, 18 students)
• Networking, speech, human centered computing
• Millennium Project (15 faculty)
• 1000 processors in campus-wide parallel
  computing facility
• Gigascale Silicon Research Center (GSRC) (23
  faculty, 60 students)
• Design tools for sub-micron silicon technology

104
CITRIS-Affiliated Research Activities(continued)

• Center for Hybrid Embedded Systems Software
  (CHESS)
• New 13M NSF Center
• Berkeley Institute of Design (BID) (10 faculty)
• New center to study design of SW, products,
  living spaces
• EECS, ME, Haas, SIMS, IEOR, CDV, CED, Art
  Practice
• Center for Image Processing and Integrated
  Computing (CIPIC) (8 faculty, 50 students)
  (UCD)
• Large scale data visualization

105
Applications-Related Current Activities

• Partners for Advanced Transit and Highways, PATH
  (20 faculty, 70 students UC,
  Caltrans, other universities)
• Technology to improve transportation in
  California
• Pacific Earthquake Engineering Research Center,
  PEER ( 25 faculty, 15 students 9
  universities),
• Identify and reduce earthquake risks
• Berkeley Seismological Laboratory (15 faculty, 14
  students)
• Runs a regional seismological monitoring system
• Studies, provides earthquake data to governments.
• National Center of Excellence in Aviation
  Operations Research, NEXTOR (6 faculty, 12
  students),
• Studies complex airport and air traffic systems.

106
Applications-Related Current Activities(continued
)

• Center for the Built Environment (CBE) (19
  faculty/staff)
• New building technologies and design techniques
• Lawrence Berkeley National Laboratory (LBNL)
• National Energy Research Supercomputing Center
  (NERSC)
• Supercomputer Center
• Environmental Energy Technologies (EET)
• Better energy-saving technologies, reduced
  environmental impact

107
Future Steps

• Build testbeds
• Training of students
• Deepen interaction with industrial and
  government partners
• Intellectual Property
• Put the social into CITRIS
• Write proposals
• CITRIS Web site
• www.citris.berkeley.edu
• This talk
• www.cs.berkeley/demmel /CITRIS_Overview_Feb03.ppt
  /