Grid Research in China and Vega Grid at ICT - IBM · 9/3/2003 · Grid Research in China and Vega...
Transcript of Grid Research in China and Vega Grid at ICT - IBM · 9/3/2003 · Grid Research in China and Vega...
Grid Research in ChinaGrid Research in Chinaand Vega Grid at ICTand Vega Grid at ICT
Zhiwei XuInstitute of Computing Technology (ICT)
Chinese Academy of Scienceswww.ict.ac.cn
September 3, 2003
Contents
• HPC/Grid market in China is looking up• Why does China need Grid?• Grid is a unique innovation opportunity• The Vega Grid project at ICT• Conclusions
• www.grid.org.cn
Computer Market Trend in China
1
10
100
1996
199719
9819
99
200020
01
200220
0320
04
2005
Hardware Software Services Total
2003 2004 2005 Annual Growth RateHardware 30.19 37.17 44.97 19.6%Software 5.56 7.05 9.04 27.3%Services 7.73 9.94 12.94 27.7%Total 43.48 54.16 66.95 22.0%
Trend: emphasizing integration, sharing, collaboration
Case 2: Grid Projects in China• 1999-2000
– Only 1 project: 863国家高性能计算环境
• 2002-2005– MOST: China National Grid (CNGrid)– CAS: e-Science Data Grid– MoE: ChinaGrid– Grid plans in Beijing and Shanghai– The NSFC Grid Project– Total government funding: 450 million RMB
• Japan National Research Grid: 100M US$• Taiwan Knowledge Innovation Grid: 30M US$
The China National Grid Project (CNGrid, 2002-2005)
• Grid Enabling Clusters (>4 Tflop/s)• Grid Nodes (8-10 Nodes, Total 6-10 Tflop/s)
– HKU node• Grid Software (Grid OS, User Environment)• Grid Applications (resource integration and sharing)
– Environment (Land Resources, Forestry, Pollution Control)– Industries (Aerospace, Automobile)– Service (Weather Forecasting)– Science (Bioinformatics, Drug Discovery, Basic Research)
ApplicationGrids
Grid Software
GridResources
Grid System Software
DevelopmentEnvironment
UseEnvironment
Science Environment Manufacture Service
HPC Equipment Data Apps
Internet
China National Grid ProjectChina National Grid Project
Sequencing at Beijing Genomics Institute• Chinese Academy of Sciences announced the discovery of
“accurate” sequence of rice genome on Dec. 13, 2002:– Map of distribution and locations of rice genome on chromosomes– Accurate locations of 97% of rice genes on 12 chromosomes– Accuracy of mapping > 99.99%– Heredity marks for breeding identified
• More HPC systems were used (altogether 6 systems)– Dawning 3000 and Dawning 2000– Sunwei system– Sun Microsystems machine– SGI system
Protein Targets
Chemicals Libraries
High Throughput Screening
Drug-Effect Computing
Dynamics Simulation
Tests
Synthesis
ACD3D 300KACD SC 1800KMDDR 100K
Drug Discovery atShanghai Institute of Materia Medica, CAS Pre-drug
chemicals
Challenges for China Railways:integration, complexity, cost
Integrated Investment/Asset Application
SPARC
Oracle
Solaris
IA-32
MySql
Linux Power4
WebSphere
AIX
VLIW
GIS
HP-UXAMD
SQL
Windows
Railway Investment Application
Railway Asset Application
• Large and uncertain quantities of data, data types, systems, users• Heterogeneity of hardware, OS, middleware, applications• Widely distributed with different administrative domains• Dynamic changes of user demands, system structure, components
Challenges for Ministry of Forestry
Organizing, managing, integration, and sharing of data that are
•Multi-source
•Multi-type
•Distributed
•Dynamic
Research Design Manufacture Test Fly Use & Maintain
Current resource level of an individual AVICS-II company
Requirements at Designer
Requirement at Plant
HW & SW Resource Usage Patterns in Aviation Products Life Cycle
The Vega Grid Project at Institute of Computing Technology (ICT)
Chinese Academy of Sciences
• Research team > 70 people (plus 100 Dawning people)• VEGA Service Grid
– Versatile Services for applications• Providing minimum common services
– Enabling Intelligence for reducing cost• Automatic, self-aware, dynamic, interactive • Providing minimum common supports
– Global Uniformity for ease of use• Connectivity, single system image, interoperability
– Autonomous Control for management• Autonomous,user-centric, open architecture
ICT Vega Grid Objectives• Scientific Impact• Economic Impact• Social Impact• International Awareness: Engaging the world
• Impact examples by Dawning Superservers• Do we have opportunities?• Can we generate such impacts?• How to do it?
Sequencing at Beijing Genomics Institute• Dawning 3000 superserver announced (2001.3)• Draft sequence of rice genome discovered (2002.4)
Reported in April 5, 2002 issue of Science
Economic (Industrial) Impact• Founder of Dawning Information Industries Co.
– Founded in 1995, 50 people– Total capital: about US$9 million
• Dawning Now– 500 people– Listed in Hong Kong Stock Exchange– Total capital: about US$100 million– 2001 Dawning Servers Revenue: US$45 million
• No. 1 among domestic high-end server companies
– 1000 Dawning DC1700 clusters shipped in 2002-2003
HPC Companies (2002)• Top 3 domestic companies for low-end servers
– Legend (spun off from ICT in 1981)– LangChao– Dawning (spun off from ICT in 1995)
• Top 3 domestic companies for HPC servers– Dawning (entered HPC market in 1995)– Legend (entered HPC market in 2001)– LangChao (entered HPC market in 2002)
• There are many other HPC vendors now
Social Impacts• Dawning 3000 and Rice Genome were listed by mass media in
10 best moments of Science & Tech in China (1997-2002)Others include the launch/return of space ship God-Ark
Established initial contacts with the world’s HPC and grid community
Visiting a China grid node by Gordon Bell, Jack Dongarra, Ian Foster, Tom Sterling, et al (May 2000)
GGG(Great Global Grid)Forbes ASAP
HP Utility Computing IBM On-Demand ComputingMicrosoft.NetSun ONE/N1
Computing and Data Grid
TeraGridIPGGIGASCI GridData Grid
Information and Knowledge Grid
Semantic WebKnowledgeManagementOntologyInformationPlatform
Business Grid
CDN
RTEC
Web Service
Other Grid Models
P2PParasitic Computing
The Great Global Grid View
Will increase IT market from $1T in 2000 to $20T in 2020
Grid = Innovation OpportunityGrid = Innovation Opportunity
Apps
Inter-face
OS
HW
Office, Browser,Games, Media
Command Line, Windows, Web;
Basic, C, Java
Windows,MacOS, Linux
IBM ArchitectureIntel Processor
PCGrid Applications
Office.Net$9B =〉 $50B
Grid User InterfaceTask-Based Interface
Grid OS.Net to Blackcomb
Grid HardwareIntel next-generation
CPUs for grid
GridKnowledge GridInformation Grid
Computing/Data Grid
Vega GridUser Interface(GSML suite)
Vega GOS
Dawning 4000Grid client
Grid “Router”
Vega
AsiaAsia--Pacific Countries Roles Pacific Countries Roles in Three IT Wavesin Three IT Waves
Internet Web GridFirst Prototype 1969.10.1 1980-1989 1998
First Testbed 1970 1990.12 1999
First Standards 1994(URI)1996(HTTP)
N/A
Std Documents by 2002
>3200 RFC’s >50
China Participation 1(1996.3) 0 ??
N/A
1969(IMP)1974(TCP/IP)
Journal of Grid Computing (started in 2003) Editorial Board
David Abramson (Australia), Satoshi Matsuoka (Japan), Satoshi Sekiguchi (Japan), Zhiwei Xu (China)
A/P
Franck Cappello (France), Jon Crowcroft (UK), FabrizioGagliardi (CERN), Tony Hey (UK), Peter Kacsuk (Poland, EIC),Domenico Laforenza (Italy), Jarek Nabrzyski (Poland), Alexander Reinefeld (Germany), Ed Seidel (Germany)
Europe
Ruth Aydt (Illinois), Fran Berman (San Diego), Mani Chandy(CalTec), Andrew Chien (San Diego), David Culler (Berkeley), Jack Dongarra (Tennessee), Ian Foster (Argonne, EIC),Jim Gray (Microsoft), Bill Johnston (Livermore), Ken Kennedy (Rice), Carl Kesselman (USC), Miron Livny (Wisconsin),Paul Messina (CalTech), Peter Steenkiste (CMU)
US
Participants at HPC/Grid Meetings
• Participants at Supercomputing 2002– Total registered participants 7200– Participants from Japan >200– Participants from China (mainland) 3
• Participants at Global Grid Forum 2002– Total registered participants 900– Participants from Japan >50– Participants from China (mainland) 1
God-sonµP Chip
龙芯
DawningHPC
曙光
VegaGrid
织女星HardwareSystem
Applications
SystemSoftware
Operation& Services
HardwareComponent
System Site Wide AreaSpace
Layer
Vega Grid is a ICT BrandVega Grid is a ICT Brand
ICT’s 3 Brands (2002-2005)
Vega Grid is Vega Grid is ““Umbrella HandleUmbrella Handle””
Vega Knowledge GridVega Information Grid
GSMLSoftware Suite
Vega Grid OS
ApplicationLayer
InterfaceLayer
System SoftwareLayer
ResourceLayer
Computing
DowningClusters
SAR
Storage
Mercury
Software
Bio.Info.
Data Mining
Multimedia
Contents
Know. Base
Security
Ontology
Network
Ipv6
Sys.Testing
Grid Arch.
Grid Router
Terminal
NC
Vega PG
God-sonµP Chip
Vega on Vega ICT on ICT
Grid Enabling: Get on the Grid
Pr ogr amLanguageVega GSML
Appl i cat i onVega- I G
Dat abaseVega- KG
CPUDawni ng
4000
OSVega GOS
St or ageMer cur y Gr i d
St or age Ser ver
NC Vega- PG
Mai nBoar dGr i d Ar ch.
A Grid Computer Model
What is the Processor?What is the Memory?What is the I/O?What is the Instruction Set?What is the OS?What is the Programming Language?What are the algorithms?
Vega Knowledge Grid (Vega-KG)
• The Internet enables global file transmission and connection of distributed hardware resources
• The World Wide Web connects globally distributed web page resources
• The Knowledge Grid is to enable intelligent service by connecting globally distributed knowledge resources
• http://kg.ict.ac.cn
Vega-KG: Publications
• The Vega-KG group has published 16 papers in international periodicals in 2003– Communications of the ACM– IEEE Intelligent Systems– IEEE Computing in Science and Engineering– Decision Support Systems – Journal of Systems and Software– Information and Management– Future Generation Computer Systems
Dawning 4000 (2004, 4-10 Tflop/s)Grid Enabling Clusters
Dagger 曙光天剑
Grid partsGrid router
Grid terminal
The Vega Grid GSML SoftwareClient Side
GSML “Browser”
grid://……
GSRP
Server Side
GSML Server
Other Servers
GCP and GCSP
Grid Operating System(e.g., Globus, OGSA, WebSphere, .Net)
Grid Resources
GSRL:Grid Service Markup LanguageGSRP:Grid Service Request ProtocolGCP: Grid Computing ProtocolGCSP:Grid Common Service Platform
Grid Community
GSML
Node2: B
Node4: D
Node1:A
Node3: C
GSML and Grid CommunityGSML Page Community Grid Resources
Beijing
Shanghai
Wuhan
Xi’an
BJDist
SH
XAWH
105
Invest
32
24167
216
Asset
342
657243
Verify
Annual investment plan mustbe sent to XYZ by 12/01.
Browser,Composer
Server,Mapper
Tech StaffSecretaryExecutive
Relation to OGSA/Relation to OGSA/GlobusGlobus,etc.,etc.
Grid Resources (computers, networks, storage, software, data, etc.)
Globus Toolkit 3.0 (GT3)
GT3 data servicesGT3 base services
GT3 core
Hosting Environment(C, J2EE, .NET, …)
GridService Other servicesService data elements
Implementation
Web Services
XMLSOAPUDDIWSDL
Environment
Vega Grid Software
OGSA/Globus Web Service Vega
GSML Software
VGOS GOS
Concluding Remarks
• Grid technology will reach the mass adoption stage in the not so distant future– Probably in 10-15 years– There are already startup companies in China
• One of the top markets will be China• Grid offers innovation opportunities• We must contribute to the world’s Grid research• China and ICT are keen to cooperate with
colleagues in the world, especially in Hong Kong