Why we need the

1
Why we need the Grid the successor to the WWW

• IOP Talk
• Dr Paul Jeffreys
• Particle Physics Department
• CLRC - Rutherford Appleton Laboratory
• p.jeffreys_at_rl.ac.uk

2
Outline

• e-Science and the Grid
• The Grid Vision
• What are Grids ?
• Grid vs web
• Challenges
• Applications
• Technical
• Example Projects
• LHC Requirements
• DataGrid
• Conclusions

3
e-Science and the Grid
4
e-Science and the Grid

• e-Science
• means science increasingly done through
  distributed global collaborations enabled by the
  Internet, using very large data collections,
  terascale computing resources and high
  performance visualisation
• Grid ...
• the word Grid is chosen by analogy with the
  electric power grid, which provides pervasive
  access to power and, like the computer and a
  small number of other advances, has had a
  dramatic impact on human capabilities and
  society. We believe that by providing pervasive,
  dependable, consistent and inexpensive access to
  advanced computational capabilities, databases,
  sensors and people, computational grids will have
  a similar transforming effect, allowing new
  classes of applications to emerge.
• Web??
• Foster Kesselman,
  1999

5
Front page FT, 7th Mar 2000
6
Who will use the Grid ?

• Computational scientists engineers
• real-time visualisation, rapid turn-round
  simulation evaluation
• Experimental scientists
• remote instrumentationsupercomputers, advanced
  visualisation
• Collaborations
• high-end VC, distributed resources (CPU, data,
  people)
• Corporations
• global enterprises, virtual environments, CAD,...
• Environmentalists
• ozone depletion, climate change, pollution--gt
  coupled models, knowledge databases
• Training education
• virtual lecture rooms - distributed classes

7
What does this all mean ?

• Enabling collaboration of dispersed communities
• Enable large scale applications (10,000 CPU,
  pipelines)
• Transparent access to high-end resources
• Uniform look feel to wide range of resources
• High availability and fault tolerance
• High performance / throughput
• Location independence of resources
• Span multiple administrative domains
• Scalability of resources
• Respect for local implementations policies

8
and what Grids are not !...

• NOT free
• organisations place LARGE resources on Grid
• NO good for close coupled computation
• e.g. CFD calculations
• NOT here yet !
• Grids built by collaborative effort to enable
  collaboration
• incrementally built no Big Bang
• many good ideas exist some implemented
• very early days of Grid integration

9
The Vision

• What are Grids ?

10
Jack Dongarras Vision
11
JD Concomitant Issues
12
JD Pictorial View
13
Guofei Jiangs Vision
14
Promise of ubiquitous computing?

• Talk from RAL.
• Download from http//www.escience.clrc.ac.uk/!!
• Global virtual computer
• Grid middleware as the OS of this computer
• Accessible via a plug on the wall
• Dependable - predictable, sustained performance
• Consistent - standard interfaces, services and
  operating parameters
• Pervasive - widely available via controlled
  access
• Inexpensive - perceived cost relative to
  alternative solutions We will
  eventually take the Grid for granted

15
Where the Grid is coming from

• Work over the last decade on several distinct
  topics
• Metacomputing - combining distributed
  heterogeneous computing resources on a single
  problem
• Data Archiving - building collections of data on
  specific topics with open metadata catalogues and
  well-documented data formats
• Collaborative Working - network-based facilities
  for distributed, concurrent working on shared
  information
• Data visualization - large-scale immersive
  facilities for 3D visual interaction with data
  coupled to computational analysis
• Network-based information systems - WAIS, Gopher,
  WWW
• Instrumentation control - remote control of
  experimental equipment and real time data
  gathering systems
• A consistent way of combining many types of
  computing resource located anywhere to solve
  complex problems

16
The Web vs The Grid

• The Web supports wide area data/information
  location and retrieval
• The Grid supports complete process initiation and
  execution including any necessary data location
  and retrieval
• offers the potential to carry out significantly
  large tasks
• opens up new capabilities for knowledge
  generation
• Currently Grid tools provide a relatively low
  level of operational control
• higher level tools will be developed to automate
  low level processes
• agent technology will eventually support real
  time dynamic process optimisation

17
NASA IPG Motivation

• NASA Information Power Grid
• Large-scale science and engineering is done
  through the interaction of people, heterogeneous
  computing resources, information systems, and
  instruments, all of which are geographically and
  organizationally dispersed.
• The overall motivation for Grids is to
  facilitate the routine interactions of these
  resources in order to support large-scale science
  and engineering.
• Motivation
• Many facilities are moving toward making
  resources available on the Grid
• The Information Power Grid is NASAs push for a
  persistent, secure, robust implementation of the
  Grid

18
IPG Case Study 1- Online Instrumentation
Unitary Plan Wind Tunnel
Multi-source Data Analysis
desktop VR clients with shared controls
real-time collection
archival storage
Computer simulations
19
IPG Case Study 2- Distributed Supercomputing

• OVERFLOW-D2 with Latency Tolerant Algorithm Mods
• Grid mechanisms for
• Resource allocation
• Distributed startup
• I/O and configuration
• Advance Reservation
• 13.5 Million Grid Points over computers at 3 or
  more sites

LaRC Origin
ANL Origin
GRC Origin
ARC Origin
20
IPG Case Study 3- Collaborative Engineering
Boeing Rapid Response Center
SFO Hangars
United MOC
Remote Displays
Digital White Board
Virtual Iron Bird
Remote Displays
United Maintenance Systems
Line Mechanic
Digital White Board
Digital White Board
Wireless Bridge
Wireless Digital - Video - Audio - Non-
Destructive Imaging - Sketchpad - Portable
Maintenance Aid
CAD/CAE Table
CAD/CAE Table
Boeing Maintenance Systems
CAD/CAE Printer
CAD/CAE Printer
CAD/CAE Printer
NASA Ames Aviation ExtraNet/ IPG
21
IPG Case Study 6- Data Grid

• Numerous data sources generating PB/yr
• Simulation 100-1000 MB/s
• Instruments Satellites 100 MB/s
• Larger, more distributed user communities
• Data are increasingly community resources
• Data analysis is increasingly multidisciplinary

22
CERN The Grid

• Dependable, consistent, pervasive access to
  high-end resources
• Dependable
• provides performance and functionality guarantees
• Consistent
• uniform interfaces to a wide variety of resources
• Pervasive
• ability to plug in from anywhere

23
CERN The Grid from a Services View

E.g.,

24
Challenges

• Application
• Technical

25
Science-driven Grid applications

• Environmental science
• coupled atmosphere and ocean simulations with
  long simulated timescales at high resolution
• Biological science
• multiple protein folding simulations to generate
  statistically valid models of complex molecules
• Astronomy - Virtual Observatory
• searching across many instrument-specific data
  archives to study a new class of object at all
  wavelengths
• Materials science
• combining and analysing data from different
  experimental facilities to derive the structure
  of complex new materials

26
Application Challenges

• Computational modelling, data analysis by
  dispersed communities, multi-disciplinary
  simulations
• aviation, HEP data analysis, climate modelling
• whole system simulation (e.g. aircraft)
• whole living cell simulation
• Online instrumentation access real-time
  analysis
• national facilities (e.g. synchrotron light
  source at ANL)
• Shared data archives
• EO data, genome data
• Collaborative visualisation analysis
• shared virtual worlds...

27
Issues

• Collaborating but dispersed research communities
• world-wide ?
• Heterogeneous resources
• desktops to supercomputers
• Multi-disciplinary
• Coupled simulation models/codes
• Data archives, curation
• Remote access to large resources
• Fast turn-round for online instruments

28
Technical Challenges

• Cross-admin domain, multi-national
• security, access policies, reimbursement
• no central control
• Resource stability
• characteristics change over time and location
• Complex distributed applications
• co-allocation, advance reservation
• optimisations (e.g. caches)
• Guaranteed end-to-end performance
• heterogeneity, fault-tolerance
• Hidden complexity

29
Technical Details - 1

• Security - a matter of trust
• map global to local domain PKI based
  certificates
• distinguish authentication from authorisation
• Data management
• caching, high throughput transfers, metadata,
  objects
• Performance
• grid view, application view
• multi-host, multi-site, non-repeatable, archives
• Workload
• multi-site, forecasting, resource/job description
  (ClassAds)
• advance resource reservation, co-allocation

30
Technical Details - 2

• Computing Fabric
• scalability, fault tolerance
• concept of service .vs. system (cluster)
• cluster hiding
• Information Services
• resource discovery
• metadata catalogue

31
Example Projects
32
SETI - Searching for Life

• Arecibo telescope in Puerto Rico
• Screensaver
• Home page in33 languages

• http//setiathome.ssl.berkeley.edu/

33
EntropiaTM

• Brokerage Free download use idle-cycles
• For-Profit Not-For-Profit Philanthropic
  Membership
• Certified project code
• Great Internet Mersenne Prime Search
• Entropia I - concept proof
• 124,380 machines, 302,449 tasks
• Fight AIDS_at_Home
• Entropia 2000 - production
• 5,589 machines, 46,716 tasks
• Future projects
• environmental, economic, scientific,
  mathematical, entertainment, product design

34
High Throughput Computing -Condor-

• HPC delivers large number cycle in bursts
• HTC delivers sustained cycles over a long period
• Job management
• Both traditional clusters idle workstations
• Heterogeneous
• UNIX
• Porting to WNT
• Sensitive to workstation use
• Checkpointing job migration
• Resource brokerage - ClassAds

35
The Globus Project

• Basic research
• resource management, storage, security,networking
  , QoS, policy, etc.

• Toolkit - bag of services, NOT an integrated
  solution
• Information Service (MDS) - GRIS/GIIS
• Remote file management - GASS
• Process monitoring - HBM
• Executable management - GEM
• Resource management - GRAM
• Security - GSI
• Many users of Globusincreasing

36
LHC Experiments Requirements
37
LHC Computing Challenge

• Data written to tape 5 Petabytes/Year (1 PB
  1015 Bytes)
• 0.1 to 1 Exabyte (1 EB 1018
  Bytes) (2010) (2020 ?) Total for the
  LHC Experiments

• Higgs
• New Particles
• Quark-Gluon Plasma
• CP Violation

38
An LHC Collaboration

• 1850 Physicists
• 150 Institutes
• 34 countries

39
CPU estimation
Capacity that can purchased for the value of the
equipment present in 2000
Non-LHC
10K SI951200 processors
LHC
technology-price curve (40 annual price
improvement)
40
Disk estimation
Non-LHC
LHC
technology-price curve (40 annual price
improvement)
41
Long term tape estimates
42
Funding

• Requirements growing faster than Moores law
• CERNs overall budget is fixed

Estimated cost of facility at CERN 30 of
offline requirements
Budget level in 2000 for all physics data
handling
assumes physics in July 2005, rapid ramp-up of
luminosity
43
Regional Computing Centres

• Exploit established computing expertise
  infrastructure
• In national labs, universities
• National funding
• Reduce dependence on links to CERN
• Active data available nearby maintained through
  a fat, fast, reliable network link
• Devolve control over resource allocation
• national interests?
• regional interests?
• at the expense of physics interests?

44
LHC Data Access Patterns
Access Rates (aggregate, average) 100 Mbytes/s
(2-5 physicists) 500 Mbytes/s (5-10
physicists) 1000 Mbytes/s (50 physicists) 2000
Mbytes/s (150 physicists)
Typical particle physics experiment in 2000-2005
One year of acquisition and analysis of data
Raw Data 1000 Tbytes
Reco-V1 1000 Tbytes
Reco-V2 1000 Tbytes
ESD-V1.1 100 Tbytes
ESD-V1.2 100 Tbytes
ESD-V2.1 100 Tbytes
ESD-V2.2 100 Tbytes
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
AOD 10 TB
45
Regional Centres - a Multi-Tier Model
46
More realistically - a Grid Topology
47
The DataGrid Project

• 21 partners, 3 years, EU-sponsored, OpenSource
• Middleware
• workload management
• data management
• performance monitoring
• compute, storage network fabrics
• Testbed
• link national PP Grids together
• EO Bioscience workpackages
• Production environment for LHC simulation
• Link with GriPhyN, PPDG in US

48
DataGrid Testbed sites
Acknowledgment A.Ghiselli
49
DataGrid Summary

• A grand vision !!
• Much technology still to develop
• Significant challenge to break down barriers
  between admin domains
• Built through collaboration to enable
  collaboration
• Will affect many disciplines (...across
  discipline ?)
• Significant growth in number of Grid projects
• EU UK government support

50
The LHC Challenge - Summary

• Scalability ? cost ? management
• Thousands of processors, thousands of disks,
  PetaBytes of data, Terabits/second of I/O b/w
• Wide-area distribution
• WANs are only and will only be 1 of LANs
• Distribute, replicate, cache, synchronise the
  data
• Multiple ownership, policies, .
• Integration of this amorphous collection of
  Regional Centres ..
• .. with some attempt at optimisation
• Adaptability
• We shall only know how analysis is done once the
  data arrives

51
CONCLUSIONS

• The overall Grid Vision is to facilitate the
  routine interactions of heterogeneous computing
  resources, information systems, and instruments,
  all of which are geographically and
  organizationally dispersed, in order to support
  large-scale science and engineering
• (NASA paraphrase)
• To meet the computing demands of the LHC
  experiments, a distributed computing model must
  be embraced, the one foreseen is based on a
  multi-tier model
• (Required for financial and sociological reasons)
• Implementing a Grid for the LHC will present
  new computing challenges
• The Grid may lead to a revolution as profound as
  the World Wide Web
• Science will never be the same again DGRC,
  Sept 2000