Download - Grid Computing in the Industry
Grid Computing in the IndustryNorth Carolina Statewide Grid
Duke University, Nov 2, 2004
Wolfgang Gentzsch Managing Director MCNC Grid Computing and Networking Services
This Lecture aims at complementing your background in distributed computing technologies and developments, providing an actual overview of Grid Computing, and its applications and benefits for Education, Industry and Economy.
NC BioGrid
MCNC Enterprise Grid
NC Statewide Grid Initiative
My current Background:MCNC & The Grid Infrastructure for North Carolina
NCREN North Carolina Research & Education Network
MCNCMicroelectronics Center of North Carolina- Founded in 1980- Independent, private, non-profit organization- Operates NCREN since 1984- Past: Supercomputer Center for universities- Now: Grid Service Provider, offering Video, Network, Grid, and Datacenter Services- 50+ employees
The Industry’s View of a Grid :
An IT Utility on a . . .
Grid Middleware (the glue)Managing . . .
Networked DistributedResources
Why Should We Care about Grids ?
“ It’s the next big thing ! “Grid technologies advance Science and Education in that we can do things which haven’t been possible before.
Grid infrastructure attracts and enables new businesses and creates new jobs, especially in today's rural areas.
Grids make us more competitive by better utilizing resources, bringing results ‘to market’ faster, and delivering with higher quality.
• Access: transparent, remote, secure, wireless• Sharing: enable collaboration over the network• Failover: migrate/restart applications automatically• On Demand: get resources, when you need them• Productivity: more work done in shorter time• Virtualization: access compute services, not servers• Heterogeneity: platforms, OSs, devices, software• Resource Utilization: increase from 20% to 80+%• Virtual Organizations: build & dismantle on the fly
Grid BenefitsDepartment, Enterprise, and Global Grids
However, There is Still a Long Way to Go ! Grids are over-hyped: currently, they promise much more than they can really offer. Grid technology is far from mature and complete. Grid standards are (mostly) still missing. Grids are very complex IT infrastructures. Grids bring new challenges: sharing resources, loosing direct control, security, intellectual property, legal, social, political issues . . .
Grid Architecture, Technology, Standards
Architectural Requirements• Availability
– Downtime impact– Individual jobs– Maintenance windows
• Scalability– Growth 1-3-5 years– Scaling strategy and – Response to peak loads– Technology refresh, evolution
• Manageability– Skill set / workload of SA– Expected stability– Code management, software
distribution• Security
– User authentication– Internet access– Data Security requirements
● Data Distribution– Location, volume, refresh– Security of data
● Usability– Administrative Skill set and – Client environment– Psychological factors
important● Operations Management
– 100s CPUs / SA– Resources added in large
blocks– Change control is critical
The Globus Toolkit and the Open Grid Services Architecture (OGSA)
• Integration of Grid technologies and Web Services• OGSA defines a “Grid Service”• In terms of WSDL interfaces, defines mechanisms:
creating distributed systemslifetime managementchange management
credential managementnotification
NotificationAuthorizationService CreationService RegistryManageabilityConcurrency Hosting Env/runtime: C, J2EE
Systems Infrastructure
Grid Service ... other interfaces
ServiceData
Element
ServiceData
Element
ServiceData
Element
Implementation / Interfaces
Global Grid Forum (GGF)• Community-driven set of working groups that
are developing standards and best practices for distributed computing ("Grids" and "Metacomputing") efforts
• Formed in 2001 (merger of US, APAC, Euro efforts)
• Open Grid Services Architecture (OGSA)
Enterprise Grid Alliance (EGA)
• Industry-driven consortium to implement standards in industry products and make them interoperable
• Founding members: Oracle, Sun, Fujitsu,
• ‘Enforce’ complementary position in pushing GGF into the academic corner
The Changing Grid Landscape
HPTC Grid
Web Svcs
Sys Mgmt
Public SOA
Technologies
Market making / Productization
Enabled Business Models
OnDemand
Utility Computing
“Compute Powerfrom the wall socket”Outsourcing/Hosting
Software as a Svc
Developer View: Write here Run anywhere Service Composition
IBM
EnterpriseIT MgmtProducts
Sun
HPIBM
. . .
Details see backup slides
Standards
Cluster Grid Implementation, Example
Access tier
Management tier
Compute tier
Login hostSGE submit host
Sun Fire V100 server
SGE admin hostSunMC console
Sun Blade 100 system
SGE masterSunMC server
Sun Fire 280R server
Install serverNFS server
Backup server
Sun Fire 280R server
Large SMP
Sun Fire V880 server
Large SMP
Sun Fire V880 server
Large SMP
Sun Fire 6800 server
with 2 CPUsLinux system
with 2 CPUsLinux system
with 2 CPUsLinux system
with 2 CPUsLinux system
with 2 CPUsLinux system
Myrinet
Department Grids
Solaris Servers
Workstations(Linux or Solaris)
Linux Servers
Internet
Browser to CGM(Remote Server Setup & Configuration)
Auto Download of Modules
Auto OS DeploymentGrid Installation/Mgmt
Centralized Server Mgmt
Compute Grid Manager
Enterprise Grid Reference Architecture
Servers,Blades,& VIZ
Workstations
Grid Manager
Linux Racks
Optional Control Network (Gbit-E)
Data Network (Gbit-E)
HA NFS Scalable QFS/NFSNAS/NFS
Myrinet
Myrinet Myrinet
Sun Fire Link
Gbit-E switch Gbit-E switch
V880 QFS/NFS Server V880 QFS/NFS Server
FC Switch
V240 / V880 NFSV240 / V880 NFS
Gbit-E switch
Simple NFS
V240 / V880 NFS
Gbit-E switch Gbit-E switch
V240 / V880 NFS
SunRay Access Browser Accessvia GEP
Workstation Access
Department 1 Department 2
Department 3
Department resource accessCampus wide resource demand
Project A
Team B
Contractor X
Project C
User 1
User 2Department 4Department 5
Grid Enterprise Policy Management
Clear definition and documentation of prioritiesand policies is key to Grid resource management
Globus & Avaki over multiple local Grids1 Grid Engine cluster mgmnt within an admin domain & file system area2 Globus/Avaki knits together resources, handles files, binary
management, and high level resource selection
a
d e
b
f
c
g h
/
ad e
Proprietary Datac
g h
Organization A Organization B Organization C
Avaki Data Grid
b
f
Clients view
SGE
SGE
Avaki HPC
SGE
Data mapped and available through Avaki Data Grid
Applications for The Grid• Single-CPU Jobs: jobmix, many users, many serial applications,
suitable for grid (e.g in universities and research centers)
• Array Jobs: 100s/1000s of jobs, one user, one serial application, varying input parameters, suitable for grid (e.g. parameter studies in Optimization, CAE, Genomics, Finance)
• Massively Parallel Jobs: one job, one user, one parallel application, no/low communication, scalable, fine-tune for grid (time-explicit algorithms, film rendering, pattern recognition)
• Parallel Jobs: one job, one user, one parallel application, high interprocs communication, not suitable for distribution over the grid, but for parallel system in the grid (time-implicit algorithms, direct solvers, large linear algebra equation systems)
Grid Computing Today in Industry
Computing as a Utility =>> Enhancing The Grid with a Business Model
• On Demand: Get a service at your finger tip• From the Wall Socket: Don't care about the infrastructure • Metering & Billing: Pay as you go, for what you use
What's a Utility?
Like electricity, water, gas, heat, telephony
Customer Challenges
Time-to-market Pressures vs. Quality Products
Multiple, concurrent projects Projects with different schedules and
milestones Overlapping demands for resources Large and growing data sets Larger, more complex product designs Not enough throughput Low utilization of compute resources Insufficient in-house IT expertise Increasingly complex operations
management Heightened focus on budget/TCO/ROI
Critical Customer Requirements
More Compute PowerImproved Resource AccessIncreased ProductivityBetter Utilization of Existing
ResourcesReduced CostsReduced Management
Complexity
Who Uses Department and Enterprise Grids?
• Life SciencesStartup and cost efficientCustom research or limited use applicationsMulti-day application runs (BLAST)Exponential CombinationsLimited administrative staffComplementary techniques
● Electronic DesignTime to MarketFastest platforms, largest GridsLicense ManagementWell established application suiteLarge legacy investmentPlatform Ownership issues● Financial Services
Market simulationsTime IS MoneyProprietary applicationsMultiple PlatformsMultiple scenario executionNeed instant results & analysis tools
● High Performance Computing
Parallel Reservoir SimulationsGeophysical Ray TracingCustom in-house codesLarge scale, multi-platform execution
How Real is all this Stuff?
department enterprise global
vendorinterest research
activity
direction of technology adaptation
Company Example:
Sun Microsystems
Sun Microsystems & Grid Computing
• Slogan: The network is the computer• Strategy: N1, manage N computers as
1• Acquisitions
● Gridware => Sun Grid Engine, Distributed Resource Manager
● Pirus => Dynamic Storage Provisioning ● Terraspring => Automatic System Provisioning ● Centerrun => Automatic Application Provisioning
Managing Grids What we did for the big SMPs...
...we are now doing for the Network
● Resource Virtualization ● Domains, Zones● Interdomain resource mgmt● Routing● Soft cabling within the box
Solaris N1
Virtualization of Resources
StorageStorage
VerticalVertical HorizontalHorizontalStorageStorageNetworkNetwork
ComputeCompute
ServiceService11
ServiceService22
ServiceService33
Compute & Data Grid
11 22 33ServicesServices ServicesServices ServicesServices
VirtualVirtualStoreStore
VirtualVirtualComputeCompute
VirtualVirtualNetworkNetwork
Services sharing resources
N1: managing services, not servers
N1 Provisions Grid Computing
Web ServerWeb ServerWeb ServerWeb Server
AppSvr
AppSvr
AppSvrDB
DB
Grid Compute Service
DB AppSvr Web Server
The N1 Effect on EfficiencyRadical improvement in costs & uptime
• Sys Utilization 6-15%80+ %
● Time to Deploy WeeksHours
● Ports/Admin50-100
500+
● Terabytes/DBA1TB
100TB
● Availability99.9 %
99.999+ %
● Server/Admin 15-30500+
Cos
tsU
ptim
e
N1
Grid Computing
Use Cases, Deployments, Best Practices
• Leeds, York + Sheffield Universities• Deliver stable, well-managed HPC resources
supporting multi-disiplinary research• Deliver a Metropolitan Grid across the
Universities
Example: White Rose Grid in England
Maxima Titania
WRG Hardware
Maxima Snowdon Pascali
Maxima
GT2.0
GEEE
Snowdon
GT2.0
GEEE
Pascali
GT2.0
GEEE
Titania
GT2.0
GEEE
portal
White Rose GridGT2.0
Solaris
Linux Solaris
Solaris
WRG Architecture Overview
WRG Key Components• Globus Toolkit 2.4Provides a secure means for inter-campus actions
● Transferring jobs● Moving data ● Gathering information about resources
• Grid Engine Enterprise EditionManages the campus grid compute resources
● Delivers a single interface for a heterogeneous grid● Guarantees a share of campus resource for grid and local users
WRG Key Components• Grid Portal Development KitProvides a portal interface into Globus Toolkit
● Transferring jobs● Moving data ● Gathering information about resources
• MyProxy MyProxy provides a server with client-side utilities to store and retrieve
delegated X.509 credentials via the Grid Security Infrastructure (GSI).
Grid Engine Enterprise Edition, Share Policies
GT2.0
GEEE
GT2.0
GEEE
Solaris
Solaris
GT2.0
GEEE
GT2.0
GEEE
Solaris
Linux
White Rose Grid
NC BioGrid
MCNC Enterprise Grid
NC Statewide Grid Initiative
Example:Grid Service Provider MCNC & The Grid Infrastructure for North Carolina
NCREN North Carolina Research & Education Network
MCNCMicroelectronics Center of North Carolina- Founded in 1980- Independent, private, non-profit organization- Operates NCREN since 1984- Past: Supercomputer Center for universities- Now: Grid Service Provider, offering Video, Network, Grid, and Datacenter Services- 50+ employees
GCNS Grid Computing and Networking Services Mission
Advance education, innovation and economic development
throughout North Carolina by delivering next generation information technology that enables the academic, research, government
and business communities
to discover, create, share and apply knowledge.
The GridThe next IT Infrastructure of North Carolina
2002 2003 2004 2005
NC BioGrid
• Proving ground for Grid• Successful prototype apps• Catalyst for collaboration• International recognition
2002 2003 2004 2005
NC BioGrid
MCNC Enterprise Grid
• Cluster and SMP resources • Research platform for GTEC• Core component in NCGrid• Revenue generation
The GridThe next IT Infrastructure of North Carolina
2002 2003 2004 2005
NC BioGrid
MCNC Enterprise Grid
NC Grid Initiative
• State-wide partnership• Leverage lessons learned• Grid education & training resource• Enable first mover applications
The Grid The next IT Infrastructure of North Carolina
CY03CY03 CY04CY04 CY0CY055 CY06CY06
MCNC: 3-Year GCNS Grid Roadmap
Easy Access: Training, Web Courses Access Grid Node Grid Appliance QuadA
Grid Service Provider:Network, Computing, Data, Video Partner with SC Sites Build GTEC Service Portfolio Start Grid Consulting Annual GSP Workshop
Awareness Creation:Conferences, Workshops, PRMCNC Enterprise GridNC BioGridNC Statewide Grid
CiscoEPA
North Carolina’s Foundation for Grid: NCREN
4-7 MCNC-owned Clusters distributed throughout the stateLocations still under evaluation
Internet Internet
2NLR
Internet
Existing: Blend of owned and leased fiber and circuits moving toward resilient rings powered by Cisco routers
Planned: Strong focus on owned and leased fiber, Lambda, and few circuits, in resilient rings powered by Cisco routers and Wave Division Multiplexers
32-CPU SGI AltixLinux SMP Server
128-CPU IBM LinuxCluster (64 nodes) 8-TB Storage
LSF Master Job Scheduler
Grid Gatekeeper / Interactive Nodes
Global Grid Resource DB
(GIIS)
Users
Campus Grids
GCNS Enterprise GridGCNS Enterprise Grid
AvakiDataGrid
Data G
rid Access Servers
(8 total, i.e. 1 per 8 nodes)
Network, Grid and Data Center Services
NCREN
State-wide Grid
Services
Enterprise Grid
Services
Value-add Information
Systems Services
Self-serve Data Center
Services
DATA CENTERHosting & Infrastructure Grid Computing
GTEC, NLR, ANR and other Innovation Initiatives
Information Security Services
Data Archival Services
Information Assurance
DEPLOYMENT
The NC State-Wide Grid Roadmap
• 06/04: “Do-Grid-Yourself” Workshop• 07/04: Phase 1, Awareness Creation• 10/04 – 06/05: Deliver/Connect Grid Appliance Clusters to University Partners in NC• 10/04: Develop “Do-Grid-Yourself” Training Course• 10/04: Start QuadA Project: Access, Accounting, Authentication, Authorization• 12/04: Start Deliver Grid Training to Partner Univs• 01/05 – 06/05: Work with Grid Users to Port Apps• 03/05: Build Access Grid Node • 06/05: 1st NC Statewide Grid Workshop
Grid Projects Driving Grid Adoption
2000 20004 2008
MCNC Grid SP Grid ColLab Kids Grids Gaming Grid School Grids Grid-Info Grid Download-The-Grid AAAA Project Grid Training Courses Startup Grid Grid Appliance Grid Portal GGF & EGA NC Statewide Grid MCNC Enterprise Grid NC BioGridMCNC Supercomputing
Grid Vision
Our Vision :The Three Waves of Grid Computing
The Research Wave The Industry Wave The Consumer Wave
Technology, Prototypes Grid-Enabled Products Commodity Virtual Organizations Enterprise Solutions IT Utility Standards Interoperability Integration GGF, IETF, OASIS GGF, EGA, IETF, OASIS Legal, Ethical, Political Orgs
GCNS: “Awareness Creation” GCNS: “Easy Access” GCNS: “Grid Service Provider”
Grids Today - Grids in 3 - 5 Years
Focus on Research - Focus on R&D and businessCompute-oriented - Petaflops linked w/ PetabytesProprietary interfaces - Standards: GGF, OGSA, DRMAA“Mental Firewall” - Security, policies, identityDifficult to build - Standards, services, solutionsDifficult to manage - Sun N1, IBM Autonomic, HP DCDifficult to use - Grid Portals: transparent, remote, secureMany technologies - Globus Toolkit 3.x
NetworkNetworkComputerComputer
Availability
Interoperability
Affordability
Manageability
Mobility—E2E
ScalabilitySecurity
Performance
Services/Solutions/Partners
The Next 10 Years:
The Grid,the Operating System of the next Internet:
High Capacity: rich in resources High Capability: rich in options Persistent: stable infrastructure & knowledgeable workforce Evolutionary: able to adapt to new technologies & users Usable: accessible, robust, and easy-to-use Scalable: growth must be part of the design Flexible: able to support new applications Fault-tolerant: resilient to changes and errors
Grid Computing: Making the Global Infrastructure a Reality
Berman, Fox, Hey, 2003
Finally:Anyone, anywhere, anytime, any device, any data,connected to The Grid Integration of new devices, data and information sources Cell phones, PDAs, smart sensors, sensor arrays, health
monitors Devices embedded in cars, engines, roads, bridges, clothes,... Huge amount of data for real-time analysis Policies, grid economy, to maintain stability and efficiency Organizational and societal structures, to bridge political
and social boundaries
Vision: Grid, the Time Machine of 21st Century
• 19th Century Time Machine: Steam Engine– Enabled mass production, reduced time to market
• 20th Century Time Machine: Combustion Engine– Much faster from Munich to Dortmund
• 21st Century Time Machine: Grid Engine– Safer, better, new, more products & services & research– in much shorter time than today !
Grid will have great impact on Science, Business, and Society ( See examples in the following presentations )
