Grids for Data Intensive Science

advertisement
Grids for Data Intensive Science
Paul Avery
University of Florida
http://www.phys.ufl.edu/~avery/
avery@phys.ufl.edu
Texas APS Meeting
University of Texas, Brownsville
Oct. 11, 2002
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
1
Outline of Talk
Grids
and Science
Data
Grids and Data Intensive Sciences
 High
Energy Physics
 Digital Astronomy
Data
Grid Projects
Networks
and Data Grids
Summary
This talk represents only a small slice of a
fascinating, multifaceted set of research efforts
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
2
Grids and Science
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
3
The Grid Concept
Grid:
Geographically distributed computing resources
configured for coordinated use
 Fabric:
Physical resources & networks provide raw capability
 Middleware: Software ties it all together (tools, services, etc.)
Goal:
Transparent resource sharing
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
4
Fundamental Idea: Resource Sharing
Resources
for complex problems are distributed
 Advanced
scientific instruments (accelerators, telescopes, …)
 Storage and computing
 Groups of people
Communities
require access to common services
 Research
collaborations (physics, astronomy, biology, eng. …)
 Government agencies
 Health care organizations, large corporations, …
Goal
 “Virtual Organizations”
 Create
a “VO” from geographically separated components
 Make all community resources available to any VO member
 Leverage strengths at different institutions
 Add people & resources dynamically
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
5
Short Comment About “The Grid”
There
is no single “Grid” a la the Internet
 Many
Grids
Grids, each devote to different organizations
are (or soon will be)
 The
foundation on which to build secure, efficient, and fair sharing
of computing resources
Grids
are not
 Sources
of free computing
 The means to access and process Petabyte-scale data freely
without thinking about it
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
6
Proto-Grid: SETI@home
Community:
Arecibo
Over
SETI researchers + enthusiasts
radio data sent to users (250KB data chunks)
2M PCs used
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
7
More Advanced Proto-Grid:
Evaluation of AIDS Drugs
Entropia
 “DCGrid”
software
 Uses 1000s of PCs
Chief
applications
 Drug
design
 AIDS research
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
8
Some (Realistic) Grid Examples
High
energy physics
 3,000
physicists worldwide pool Petaflops of CPU resources to
analyze Petabytes of data
Climate
modeling
 Climate
scientists visualize, annotate, & analyze Terabytes of
simulation data
Biology
A
biochemist exploits 10,000 computers to screen 100,000
compounds in an hour
Engineering
A
multidisciplinary analysis in aerospace couples code and data in
four companies to design a new airframe
Many
commercial applications
From Ian Foster
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
9
Grids: Why Now?
Moore’s
law improvements in computing
 Highly
functional endsystems
Universal
wired and wireless Internet connections
 Universal
Changing
connectivity
modes of working and problem solving
 Interdisciplinary
teams
 Computation and simulation as primary tools
Network
 (Next
exponentials
slide)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
10
Network Exponentials & Collaboration
Network
(WAN) vs. computer performance
 Computer
speed doubles every 18 months
 WAN speed doubles every 12 months (revised)
 Difference = order of magnitude per 10 years
 Plus ubiquitous network connections!
1986
to 2001
 1,000
 Networks:  50,000
 Computers:
2001
to 2010?
 60
 Networks:  500
 Computers:
Scientific American (Jan-2001)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
11
Basic Grid Challenges
Overall
goal: Coordinated sharing of resources
 Resources
Many
under different administrative control
technical problems to overcome
 Authentication,
authorization, policy, auditing
 Resource discovery, access, negotiation, allocation, control
 Dynamic formation & management of Virtual Organizations
 Delivery of multiple levels of service
 Autonomic management of resources
 Failure detection & recovery
Additional
issue: lack of central control & knowledge
 Preservation
of local site autonomy
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
12
Advanced Grid Challenges: Workflow
Manage
workflow across Grid
 Balance
policy vs. instantaneous capability to complete tasks
 Balance effective resource use vs. fast turnaround for priority jobs
 Match resource usage to policy over the long term
 Goal-oriented algorithms: steering requests according to metrics
Maintain
a global view of resources and system state
 Coherent
end-to-end system monitoring
 Adaptive learning: new paradigms for execution optimization
Handle
user-Grid interactions
 Guidelines,
Build
agents
high level services & integrated user environment
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
13
Layered Grid Architecture
(Analogy to Internet Architecture)
Application
User
Managing multiple resources:
ubiquitous infrastructure services
Collective
Sharing single resources:
negotiating access, controlling use
Talking to things:
communications, security
Application
Resource
Connectivity
Transport
Internet
Fabric
Link
Controlling things locally:
Accessing, controlling resources
Internet Protocol Architecture
Specialized services:
App. specific distributed services
From Ian Foster
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
14
Globus Project and Toolkit
Globus
Project™ (UC/Argonne + USC/ISI)
 O(40)
researchers & developers
 Identify and define core protocols and services
Globus
Toolkit™ 2.0
 Reference
Globus
Toolkit used by most Data Grid projects today
 US:
GriPhyN, PPDG, TeraGrid, iVDGL, …
EU-DataGrid and national projects
 EU:
Recent
implementation of core protocols & services
progress: OGSA and web services (2002)
 OGSA:
Open Grid Software Architecture
 Applying “web services” to Grids: WSDL, SOAP, XML, …
 Keeps Grids in the commercial mainstream
 Globus ToolKit 3.0
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
15
Data Grids
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
16
Data Intensive Science: 2000-2015
Scientific
discovery increasingly driven by IT
 Computationally
intensive analyses
 Massive data collections
 Data distributed across networks of varying capability
 Geographically distributed collaboration
Dominant
 2000
 2005
 2010
 2015
factor: data growth (1 Petabyte = 1000 TB)
~0.5 Petabyte
~10 Petabytes
~100 Petabytes
~1000 Petabytes?
How to collect, manage,
access and interpret this
quantity of data?
Drives demand for “Data Grids” to handle
additional dimension of data access & movement
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
17
Data Intensive Physical Sciences
High
energy & nuclear physics
 Including
Gravity
new experiments at CERN’s Large Hadron Collider
wave searches
 LIGO,
GEO, VIRGO
Astronomy:
Digital sky surveys
 Sloan
Digital sky Survey, VISTA, other Gigapixel arrays
 “Virtual” Observatories (multi-wavelength astronomy)
Time-dependent
3-D systems (simulation & data)
 Earth
Observation, climate modeling
 Geophysics, earthquake modeling
 Fluids, aerodynamic design
 Pollutant dispersal scenarios
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
18
Data Intensive Biology and Medicine
Medical
data
 X-Ray,
mammography data, etc. (many petabytes)
 Digitizing patient records (ditto)
X-ray
crystallography
 Bright
X-Ray sources, e.g. Argonne Advanced Photon Source
Molecular
genomics and related disciplines
 Human
Genome, other genome databases
 Proteomics (protein structure, activities, …)
 Protein interactions, drug delivery
Brain
Craig Venter keynote
@SC2001
scans (1-10m, time dependent)
Virtual
Population Laboratory (proposed)
 Database
of populations, geography, transportation corridors
 Simulate likely spread of disease outbreaks
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
19
Example: High Energy Physics @ LHC
“Compact” Muon Solenoid
at the LHC (CERN)
Smithsonian
standard man
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
20
CERN LHC site
CMS
LHCb
ALICE
Texas APS Meeting (Oct. 11, 2002)
Atlas
Paul Avery
21
Collisions at LHC (2007?)
ProtonProton
Protons/bunch
Beam energy
Luminosity
2835 bunch/beam
1011
7 TeV x 7 TeV
1034 cm2s1
Bunch
Crossing rate
Every 25 nsec
Proton
Collision rate ~109 Hz
(Average ~20 Collisions/Crossing)
Parton
(quark, gluon)
l
l
Higgs
o
Z
+
e
Particle
e+
New physics rate ~ 105 Hz
e-
o
Z
jet
Texas APS Meeting (Oct. 11, 2002)
jet
e-
Selection: 1 in 1013
SUSY.....
Paul Avery
22
Data Rates: From Detector to Storage
Physics filtering
40 MHz
~1000 TB/sec
Level 1 Trigger: Special Hardware
75 GB/sec
75 KHz
Level 2 Trigger: Commodity CPUs
5 GB/sec
5 KHz
Level 3 Trigger: Commodity CPUs
100 MB/sec
100 Hz
Raw Data to storage
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
23
LHC Data Complexity
“Events”
resulting from beam-beam collisions:
 Signal
event is obscured by 20 overlapping uninteresting collisions
in same crossing
 CPU time does not scale from previous generations
2000
Texas APS Meeting (Oct. 11, 2002)
2007
Paul Avery
24
LHC: Higgs Decay into 4 muons
(+30 minimum bias events)
All charged tracks with pt > 2 GeV
Reconstructed tracks with pt > 25 GeV
109 events/sec, selectivity: 1 in 1013
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
25
LHC Computing Overview
Complexity:
Millions of individual detector channels
Scale:
PetaOps (CPU), Petabytes (Data)
Distribution:
Global distribution of people & resources
1800 Physicists
150 Institutes
32 Countries
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
26
Global LHC Data Grid
Experiment (e.g., CMS)
Tier0/( Tier1)/( Tier2) ~1:1:1
~100 MBytes/sec
Online
System
Tier 0
2.5 Gbits/sec
Tier 1
France
Italy
UK
CERN Computer
Center > 20 TIPS
USA
2.5 Gbits/sec
Tier 2
Tier2 Center
Tier2 Center
Tier2 Center
Tier2 Center
Tier2 Center
~0.6 Gbits/sec
Tier 3
InstituteInstitute Institute
~0.25TIPS
Institute
0.1 - 1 Gbits/sec
Tier 4
Physics data cache
PCs, other portals
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
27
LHC Tier2 Center (2001)
“Flat” switching topology
FEth/GEth
Switch
WAN
Router
>1 RAID
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
Tape
28
LHC Tier2 Center (2001)
“Hierarchical” switching topology
FEth Switch
FEth Switch
GEth Switch
FEth Switch
FEth Switch
WAN
Router
>1 RAID
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
Tape
29
Hardware Cost Estimates
Buy
1.4 years
1.1 years
2.1 years
1.2 years
late, but not too late: phased implementation
 R&D
Phase
2001-2004
 Implementation Phase
2004-2007
 R&D to develop capabilities and computing model itself
 Prototyping at increasing scales of capability & complexity
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
30
Example: Digital Astronomy Trends
Future dominated by detector improvements
1000
• Moore’s Law growth in CCDs
100
• Gigapixel arrays on horizon
10
• Growth in CPU/storage
tracking data volumes
1
• Investment in software critical
Glass
MPixels
0.1
1970
1975
1980
1985
1990
1995
2000
CCDs
Glass
•Total area of 3m+ telescopes in the world in m2
•Total number of CCD pixels in Mpix
•25 year growth: 30x in glass, 3000x in pixels
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
31
The Age of Mega-Surveys
Next
generation mega-surveys will change astronomy
 Top-down
design
 Large sky coverage
 Sound statistical plans
 Well controlled, uniform systematics
The
technology to store and access the data is here
 We
are riding Moore’s law
Integrating
these archives is for the whole community
 Astronomical
data mining will lead to stunning new discoveries
 “Virtual Observatory”
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
32
Virtual Observatories
Multi-wavelength astronomy,
Multiple surveys
Standards
Source Catalogs
Image Data
Specialized Data:
Information Archives:
Spectroscopy, Time Series,
Polarization
Discovery Tools:
Derived & legacy data:
NED,Simbad,ADS, etc
Visualization, Statistics
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
33
Virtual Observatory Data Challenge
Digital
representation of the sky
 All-sky
+ deep fields
 Integrated catalog and image databases
 Spectra of selected samples
Size
of the archived data
 40,000
square degrees
 Resolution < 0.1 arcsec  > 50 trillion pixels
 One band (2 bytes/pixel)
100 Terabytes
 Multi-wavelength:
500-1000 Terabytes
 Time dimension:
Many Petabytes
Large,
globally distributed database engines
 Multi-Petabyte
data size
 Thousands of queries per day, Gbyte/s I/O speed per site
 Data Grid computing infrastructure
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
34
Sloan Sky Survey Data Grid
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
35
Data Grid Projects
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
36
New Collaborative Endeavors via Grids
Fundamentally
 Old:
 New:
alters conduct of scientific research
People, resources flow inward to labs
Resources, data flow outward to universities
Strengthens
 Couples
universities
universities to data intensive science
 Couples universities to national & international labs
 Brings front-line research to students
 Exploits intellectual resources of formerly isolated schools
 Opens new opportunities for minority and women researchers
Builds
partnerships to drive new IT/science advances
 Physics
 Application
 Universities
sciences
 Fundamental
sciences
 Research Community
Texas APS Meeting (Oct. 11, 2002)





Astronomy, biology, etc.
Computer Science
Laboratories
IT infrastructure
IT industry
Paul Avery
37
Background: Major Data Grid Projects
 Particle Physics Data Grid (US, DOE)
 Data Grid applications for HENP expts.
 GriPhyN (US, NSF)
 Petascale Virtual-Data
 iVDGL (US, NSF)
 Global Grid lab
Grids
 Data
 DOE Science Grid (DOE)
 Link major DOE computing
 TeraGrid (US, NSF)
 Dist. supercomp.
sites
resources (13 TFlops)
 European Data Grid (EU,
 Data Grid technologies,
 CrossGrid (EU, EC)
 Realtime Grid tools
 DataTAG (EU, EC)
 Transatlantic network,
EC)
EU deployment
intensive expts.
 Collaborations
of
application scientists &
computer scientists
 Infrastructure
deployment
 Globus
devel. &
based
Grid applications
 Japanese Grid Project (APGrid?) (Japan)
 Grid deployment throughout Japan
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
38
GriPhyN: PetaScale Virtual-Data Grids
Production Team
Individual Investigator
Interactive User Tools
Virtual Data Tools
Request Planning &
Scheduling Tools
Resource
èResource
èManagement
Management
èServices
Services
Workgroups
~1 Petaflop
~100 Petabytes
Request Execution &
Management Tools
èSecurity
and
Security
and
èPolicy
Policy
èServices
Services
Other Grid
Services
èOther Grid
èServices
Transforms
Distributed resources
Raw data
source
Texas APS Meeting (Oct. 11, 2002)
(code, storage, CPUs,
networks)
Paul Avery
39
Major facilities, archives
Virtual Data Concept

Data request may





Compute locally
Compute remotely
Access local data
Access remote data
Regional facilities, caches
Scheduling based on



Local policies
Global policies
Cost
Texas APS Meeting (Oct. 11, 2002)
Fetch item
Paul Avery
Local facilities, caches
40
Early GriPhyN Challenge Problem:
CMS Data Reconstruction
Master Condor
job running at
Caltech
5) Secondary
reports complete
to master
Caltech
workstation
6) Master starts
reconstruction jobs
via Globus
jobmanager on
cluster
April 2001
Caltech
NCSA
Wisconsin
2) Launch secondary job on Wisconsin
pool; input files via Globus GASS
Secondary
Condor job on UW
pool
3) 100 Monte
Carlo jobs on
Wisconsin Condor
pool
9) Reconstruction
job reports
complete to master
7) GridFTP fetches
data from UniTree
4) 100 data files
transferred via
GridFTP, ~ 1 GB
each
NCSA Linux cluster
Texas APS Meeting (Oct. 11, 2002)
8) Processed
objectivity
database stored
to UniTree
Paul Avery
NCSA UniTree
- GridFTPenabled FTP
server
41
Particle Physics Data Grid



Funded by DOE MICS ($9.5M for 2001-2004)
DB replication, caching, catalogs
Practical orientation: networks, instrumentation, monitoring
Computer Science Program of Work
 CS1: Job Description Language
 CS2: Schedule and Manage Data
Processing & Placement Activities
 CS3 Monitoring and Status Reporting
 CS4 Storage Resource Management
 CS5 Reliable Replication Services
 CS6 File Transfer Services
 ….
 CS11 Grid-enabled Analysis
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
42
iVDGL: A Global Grid Laboratory
“We propose to create, operate and evaluate, over a
sustained period of time, an international research
laboratory for data-intensive science.”
From NSF proposal, 2001
International
A
A
A
A
A
U.S.
Virtual-Data Grid Laboratory
global Grid laboratory (US, EU, Asia, South America, …)
place to conduct Data Grid tests “at scale”
mechanism to create common Grid infrastructure
laboratory for other disciplines to perform Data Grid tests
focus of outreach efforts to small institutions
part funded by NSF (2001-2006)
 $13.7M
(NSF) + $2M (matching)
 UF directs this project
 International partners bring own funds
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
43
Current US-CMS Testbed (30 CPUs)
Wisconsin
Princeton
Fermilab
Caltech
UCSD
Florida
Brazil
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
44
US-iVDGL Data Grid (Dec. 2002)
SKC
LBL
Wisconsin Michigan
PSU
Fermilab
Argonne
NCSA
Caltech
Oklahoma
Indiana
Paul Avery
J. Hopkins
Hampton
FSU
Arlington
Texas APS Meeting (Oct. 11, 2002)
BNL
Vanderbilt
UCSD/SDSC
Brownsville
Boston U
UF
Tier1
Tier2
Tier3
FIU
45
iVDGL Map (2002-2003)
Surfnet
DataTAG
New partners
Brazil
T1
Russia
T1
Chile
T2
Pakistan T2
China
T2
Romania ?
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
Tier0/1 facility
Tier2 facility
Tier3 facility
10 Gbps link
2.5 Gbps link
622 Mbps link
Other link
46
TeraGrid: 13 TeraFlops, 40 Gb/s
Site Resources
26
4
Site Resources
HPSS
HPSS
24
8
External
Networks
Caltech
External
Networks
Argonne
40 Gb/s
External
Networks
Site Resources
HPSS
Texas APS Meeting (Oct. 11, 2002)
SDSC
4.1 TF
225 TB
5
NCSA/PACI
8 TF
240 TB
Paul Avery
External
Networks
Site Resources
UniTree
47
DOE Science Grid
Link
major DOE computing sites (LBNL)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
48
EU DataGrid Project
Work
Package
Work Package title
Lead
contractor
WP1
Grid Workload Management
INFN
WP2
Grid Data Management
CERN
WP3
Grid Monitoring Services
PPARC
WP4
Fabric Management
CERN
WP5
Mass Storage Management
PPARC
WP6
Integration Testbed
CNRS
WP7
Network Services
CNRS
WP8
High Energy Physics Applications
CERN
WP9
Earth Observation Science Applications
ESA
WP10
Biology Science Applications
INFN
WP11
Dissemination and Exploitation
INFN
WP12
Project Management
CERN
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
49
LHC Computing Grid Project
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
50
Need for Common Grid Infrastructure
Grid
computing sometimes compared to electric grid
 You
plug in to get a resource (CPU, storage, …)
 You don’t care where the resource is located
This analogy is more appropriate than originally intended
expresses a USA viewpoint  uniform power grid
What happens when you travel around the world?
It
Different frequencies
Different voltages
Different sockets!
60 Hz, 50 Hz
120 V, 220 V
USA, 2 pin, France, UK, etc.
Want to avoid this situation in Grid computing
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
51
Grid Coordination Efforts
Global
Grid Forum (GGF)
 www.gridforum.org
 International
forum for general Grid efforts
 Many working groups, standards definitions
 Next one in Toronto, Feb. 17-20
HICB
(High energy physics)
 Represents
HEP collaborations, primarily LHC experiments
 Joint development & deployment of Data Grid middleware
 GriPhyN, PPDG, TeraGrid, iVDGL, EU-DataGrid, LCG, DataTAG,
Crossgrid
 Common testbed, open source software model
 Several meeting so far
New
infrastructure Data Grid projects?
 Fold
into existing Grid landscape (primarily US + EU)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
52
Networks
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
53
Next Generation Networks for HENP
Rapid
access to massive data stores
 Petabytes
Balance
and beyond
of high throughput vs rapid turnaround
 Coordinate
Seamless
& manage: Computing, Data, Networks
high performance operation of WANs & LANs
 WAN:
Wide Area Network
 LAN:
Local Area Network
 Reliable, quantifiable, high performance
 Rapid access to the data and computing resources
 “Grid-enabled” data analysis, production and collaboration
Full
participation by all physicists, regardless of location
 Requires
good connectivity
 Grid-enabled software, advanced networking, collaborative tools
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
54
2.5 Gbps Backbone
201 Primary Participants
All 50 States, D.C. and Puerto Rico
75 Partner Corporations and Non-Profits
14 State Research and Education Nets
15 “GigaPoPs” Support 70% of Members
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
55
Total U.S. Internet Traffic
100 Pbps
Limit of same % GDP as Voice
10 Pbps
1 Pbps
100Tbps
New Measurements
10Tbps
1Tbps
100Gbps
Projected at 4/Year
Voice Crossover: August 2000
10Gbps
1Gbps
ARPA & NSF Data to 96
100Mbps
10Mbps
4X/Year
2.8X/Year
1Mbps
100Kbps
10Kbps
1Kbps
100 bps
10 bps
1970
1975
1980
1985
1990
1995
2000
2005
2010
U.S. Internet Data Traffic
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
56
Source: Roberts et al., 2001
Bandwidth for the US-CERN Link
Link Bandwidth (Mbps)
10000
Evolution typical
of major HENP
links 2001-2006
8000
6000
4000
2000
0
FY2001 FY2002 FY2003 FY2004 FY2005 FY2006
BW (Mbps)
310
622
1250
2500
5000
10000
2155
Mbps in 2001
622 Mbps May 2002
2.5 Gbps Research Link Summer 2002 (DataTAG)
10 Gbps Research Link in mid-2003 (DataTAG)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
57
Transatlantic Network Estimates
2001 2002 2003 2004
CMS
2005
2006
100
200
300
600
800
2500
ATLAS
50
100
300
600
800
2500
BaBar
300
600 1100 1600 2300
3000
CDF
100
300
400 2000 3000
6000
D0
400 1600 2400 3200 6400
8000
BTeV
20
40
100
200
300
500
DESY
100
180
210
240
270
300
CERN
311
622 2500 5000 10000 20000
BW
in Mbps, assuming 50% utilization
See http://gate.hep.anl.gov/lprice/TAN
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
58
All Major Links Advancing Rapidly
Next
generation 10 Gbps national network backbones
 Starting
Major
to appear in the US, Europe and Japan
transoceanic links
 Are/will
Critical
be at 2.5 - 10 Gbps in 2002-2003
path
 Remove
regional, last mile bottlenecks
 Remove compromises in network quality
 Prevent TCP/IP inefficiencies at high link speeds
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
59
U.S. Cyberinfrastructure Panel:
Draft Recommendations (4/2002)
New
initiative to revolutionize science, engineering research
 Capitalize
on new computing & communications opportunities
 Supercomputing, massive storage, networking, software,
collaboration, visualization, and human resources
 Budget estimate: incremental $650 M/year (continuing)
New
office with highly placed, credible leader
 Initiate
competitive, discipline-driven path-breaking applications
 Coordinate policy and allocations across fields and projects
 Develop middleware & other software essential to scientific research
 Manage individual computational, storage, and networking resources
at least 100x larger than individual projects or universities
Participants
 NSF
directorates, Federal agencies, international e-science
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
60
Summary
Data
Grids will qualitatively and quantitatively change the
nature of collaborations and approaches to computing
Current
Data Grid projects will provide vast experience for
new collaborations, point the way to the future
Networks
Many
must continue exponential growth
challenges during the coming transition
 New
grid projects will provide rich experience and lessons
 Difficult to predict situation even 3-5 years ahead
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
61
Grid References
 Grid Book
 www.mkp.com/grids
 Globus
 www.globus.org
 Global Grid Forum
 www.gridforum.org
 TeraGrid
 www.teragrid.org
 EU DataGrid
 www.eu-datagrid.org
 PPDG
 www.ppdg.net
 GriPhyN
 www.griphyn.org
 iVDGL
 www.ivdgl.org
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
62
More Slides
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
63
1990s Information Infrastructure
O(107) nodes
Network
Network-centric
 Simple,
fixed end systems
 Few embedded capabilities
 Few services
 No user-level quality of service
From Ian Foster
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
64
Emerging Information Infrastructure
O(1010) nodes
Application-centric
Caching
Resource
Discovery
Processing
QoS
Grid
 Heterogeneous,
mobile end-systems
 Many embedded capabilities
 Rich services
 User-level quality of service
Qualitatively different,
not just “faster and
more reliable”
From Ian Foster
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
65
Globus General Approach
Define
Applications
Grid protocols & APIs
 Protocol-mediated
access to remote resources
 Integrate and extend existing standards
Develop
reference implementation
Diverse global services
 Open
source Globus Toolkit
 Client & server SDKs, services, tools, etc.
Grid-enable
wide variety of tools
 Globus
Toolkit
 FTP, SSH, Condor, SRB, MPI, …
Learn
about real world problems
Core
services
 Deployment
 Testing
 Applications
Diverse resources
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
66
ICFA SCIC
SCIC:
Standing Committee on Interregional Connectivity
 Created
by ICFA in July 1998 in Vancouver
 Make recommendations to ICFA concerning the connectivity
between the Americas, Asia and Europe
SCIC
duties
 Monitor
traffic
 Keep track of technology developments
 Periodically review forecasts of future bandwidth needs
 Provide early warning of potential problems
 Create subcommittees when necessary
Reports:
February, July and October 2002
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
67
SCIC Details
Network
status and upgrade plans
 Bandwidth
and performance evolution
 Per country & transatlantic
Performance
Study
measurements (world overview)
specific topics
 Example:
Bulk transfer, VoIP, Collaborative Systems, QoS, Security
Identification
of problem areas
 Ideas
on how to improve, or encourage to improve
 E.g., faster links  equipment cost issues, TCP/IP scalability, etc.
Meetings
 Summary
and sub-reports available (February, May, October)
 http://www.slac.stanford.edu/grp/scs/trip/notes-icfa-dec01cottrell.html
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
68
Internet2 HENP Working Group
Mission:
Ensure the following HENP needs
 National
and international network infrastructures (end-to-end)
 Standardized tools & facilities for high performance end-to-end
monitoring and tracking
 Collaborative systems
Meet
HENP needs in a timely manner
 US
LHC and other major HENP Programs
 At-large scientific community
 Create program broadly applicable across many fields
Internet2
Working Group: Oct. 26 2001
 Co-Chairs:
S. McKee (Michigan), H. Newman (Caltech)
 http://www.internet2.edu/henp
(WG home page)
 http://www.internet2.edu/e2e
(end-to-end initiative)
Texas APS Meeting (Oct. 11, 2002)
Paul Avery
69
Download