TLC2 Infrastructure Efforts and Grid Related Research

TLC2 Infrastructure Efforts
and
Grid Related Research
Lennart Johnsson
Director TLC2
Department of Computer Science
UHCC 2nd floor
Fleming 232
SR-1, 232
10 TB
SR-3, 402G
10 TB
rackmount
Layer-3
GigE
MM
5 TB
rackmount
Layer-2 GigE
PC Picture
Layer-2 GigE
Layer-2 GigE
MM
SM
SM
Fleming 117
SM
SM
SM SR-1 1xx
“PGH 10”
RFO863609 Labels
A.
Switch in “PGH-10”
B.
Switches at Baylor/Rice
PGH 2XX
C.
D.
Switches at UHCC/PGH576
Alternate to B
MM
MM
Layer-3
1/10 GigE
1/10 GigE
Dark fiber
Baylor
10 GigE
1/10 GigE
MM
MM
Layer-3
GigE
Rice
MM
PGH 6XX
15 TB
PGH 576
SR-3 1xx
10 GigE
rackmount
PGH
216
SM
MM
Layer-2 GigE
PGH 660
E-Science: Data Gathering, Analysis, Simulation, and Collaboration
Simulated Higgs Decay
CMS
LHC
Optical Communication costs
Larry Roberts, Caspian Networks
1000
8KB
Average Price of Storage
DRAM
32KB
100
Price/MByte, Dollars
In 2010, $1 will buy enough
disk space to store
Seagate ST500
Wren II
10
1
Range
0.1
0.01
HDD
64KB
DRAM
oemprc2000aa.prz
Paper/Film
256KB Flash
512KB Flash
2MB
1MB
4MB Flash
1MB Flash
4MB
IBM0615
64MB Flash
16MB Flash
Maxt170
96 MB Flash Camera
128MB Flash
Mem.
128MB Flash
64MB
IBM0663
IBM 340 MB Microdrive
Flash
64MB
IBM 9.1GB Ultrastar
IBM 1 GB Microdrive
Seagate B'cuda4
IBM 8.1GB
Flash Projection
IBM Deskstar3
of Paper/Film
Travelstar
DataQuest 2000
IBM Deskstar4
Toshiba
Quant 4.5GB
1" HDD Projection
IBM 25GB
IBM 18.2GB Ultrastar
6.4GB
DataQuest 2000
IBM 16.8GB Deskstar
Travelstar
Seagate 8.6GB
IBM 9.1GB Ultra 2XP
IBM Deskstar 25GB
1 " HDD
IBM Deskstar 37GB
IBM Deskstar 75GXP
3.5 " HDD
0.001
1980
Flash
128KB
IBM6150
128KB Flash
512KB
Seagate ST125
1985
1990
1995
2000
2.5 " HDD
2005
2010
Year
Ed Grochowski at Almaden
10,000 Books
35 hrs of CD
Quality audio
2 min of DVD
Quality Video
Functions per chip, Mtransistors
100000
Cost of Computing
80000
DRAM, at Production
70000
DRAM, at Introduction
60000
50000
40000
Today’s most powerful
computers (the power of
10,000 PCs at a cost of
$100M) will cost a few
hundred thousand dollars
30000
20000
MPU, Cost-Performance, at
Production
MPU, Cost-Performance, at
Introduction
MPU, High-Performance, at
Production
ASIC, at Production
10000
0
1999
180 nm
2000
2001 2002
130 nm
.
2003
.
2004 2005
90 nm
.
2008 2011
2014
60 nm 40 nm 30 nm
Cost/Mtransistor
$/Mtransistor
In 2010, the compute power
of today’s top-of-the-line PC
can be found in $1 consumer
electronics
SIA Roadmap
90000
45
40
35
30
25
20
15
10
5
0
SIA Roadmap
DRAM, cost x 100, at Introduction,
$/Mtransistors
DRAM, cost x 100, at Production,
$/Mtransistors
1999 2001 2003 2005 2011
.
100 40 nm
130
180
nm
nm
nm
MPU, Cost-Performance, at
Introduction, $/Mtransistors
MPU, Cost-Performance, at
Production, $/Mtransistors
MPU, High Performance, at
Production, $/Mtransistors
GUSTO tesbed for Grid applications demonstrated at
Supercomputing97 exhibition
Tele-Microscopy
Osaka, Japan
Mark Ellisman, UCSD
www.givenimaging.com
Medicine On-line
Ê New Sensors—Israeli Video Pill
– Battery, Light, & Video Camera
– Images Stored on Hip Mounted Device
Ê Next Step—Putting You On-Line!
– Wireless Internet Transmission
– Key Metabolic and Physical Variables
Ê Post-Genomic Individualized Medicine
– Combine
• Genetic Code
• Body Data Flow
– Use Powerful AI Data Mining Techniques
www.bodymedia.com
BIOMEDICAL INFORMATICS RESEARCH
NETWORK (BIRN)
The BIRN is an NCRR initiative aimed at creating a testbed to address biomedical
researchers' need to access and analyze data at a variety of levels of aggregation located
at diverse sites throughout the country. The BIRN testbed will bring together hardware
and develop software necessary for a scalable network of databases and computational
resources. Issues of user authentication, data integrity, security, and data ownership will
also be addressed. $20M.
A NATIONAL DIGITAL MAMMOGRAPHY ARCHIVE
Digital Mammography
32.5 million mammograms/yr in the US.
Resolution 50 micron
5000 x 4000 = 20,000,000 pixels
Dynamic range: 12 bits per pixel
Uncompressed image: 40 Mbyte (16 bpp)
Four images per patient: 160 Mbytes
Total: 5 Pbytes/yr, uncompressed
Lossless compression: reduction by a factor of 2 – 3
Lossy compression: reduction by up to a factor of 80
National Ecological Observatory Network
$ 100M FY01-06 NSF support requested
Ê 10 observatories nationwide; sponsored by NSF, Archbold Biological Station and SDSC
Ê 10 geographically distributed observatories nationwide to serve as national
research platforms for integrated, cutting-edge research in field biology
Ê To enable scientists to conduct experiments on ecological systems at all levels of
biological organization – from molecular genetics to whole ecosystems, and
across scales – from seconds to geological time, and from microns to regions and
continents.
Ê Observatories will have scalable computation capabilities and will be networked
via satellite and landlines – to each other and to specialized facilities, such as
supercomputer centers.
Ê By creating one virtual installation via a cutting-edge computational network, all
http://www.sdsc.edu/NEON/
members
of the field biology research community will be able to access NEON
remotely.
Large Hadron Collider
Large Hadron Collider (LHC)
$ 80.9M FY99-04 NSF support requested
Ê CERN (Switzerland) and international Consortium
Ê Construction of two detectors of the LHC:
ATLAS (A Toroidal Large Angle
Spectrometer) and CMS (Compact Muon
Solenoid)
Ê The research, design, and prototyping of
Petascale Virtual Data Grids, which will
support the LHC as well as the SDSS
(Sloan Digital Sky Survey) and LIGO (Laser
Interferometer Gravitational-wave
Observatory), is being carried out by
GriPhyN, a multi- institutional team that
received the largest NSF ITR grant in FY00.
http://lhc.web.cern.ch/lhc/
National Data Observatory
The National Virtual Observatory (NVO) will unite astronomical databases of many
earthbound and orbital observatories, taking advantage of the latest computer technology
and data storage and analysis techniques. The goal is to maximize the potential for new
scientific insights from the data by making them available in an accessible, seamlessly
unified form to professional researchers, amateur astronomers and students. $10 M.
http://us-vo.org
Alliance98
Interactive Collaborative Virtual Environment
http://www.pdc.kth.se/projects/alliance98/
The Grid: Computational Steering
GEMSviz at iGRID 2000
Paralle ldatorc ent rum
KTH Stockholm
INET
NORDUnet
STAR TAP
APAN
Universityof Hous ton
http://www.pdc.kth.se/projects/gemsviz
NORDUnet 2000
Erik Engquist
28 Sep 00 - #17
SimDB
Simulation Data Base
Grid Application Development
Software (GrADS)
Grids – Resource Allocation
Grids – Overhead
Cactus – Migration example
Processor-DRAM Memory Gap (latency)
CPU
“Moore’s Law”
10
DRAM
1
µProc
60%/yr.
(2X/1.5yr)
Processor-Memory
Performance Gap:
(grows 50% / year)
100
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001_
_
Performance
1000
Time
DRAM
9%/yr.
(2X/10 yrs)
UHFFT Architecture
UHFFT
Library
Unparser
Library of
FFT Modules
Initialization
Routines
FFT Code
Generator
Mixed-Radix
(Cooly-Tukey)
Execution
Routines
Prime Factor
Algorithm
Utilities
Split-Radix
Algorithm
Scheduler
Key:
Optimizer
Initializer
(Algorithm Abstraction)
Fixed library code
Generated code
Code generator
Rader's
Algorithm
Performance Tuning Methodology
Input Parameters
System specifics,
User options
UHFFT Code
generator
Input Parameters
Size, dim., …
Library of
FFT modules
Initialization
Select best plan
Performance
database
Execution
Calculate one
or more FFTs
Installation
Run-time