1october, 2001 sun in scientific & engineering computing grid computing for life sciences...
TRANSCRIPT
1October , 2001
Sun in Scientific & Engineering Computing
Grid Computing for Life Sciences
Wolfgang Gentzsch Director Grid Computing
BioGrid Symposium, Singapore, October, 2001
2October , 2001
The BIG Challenges: Reality !
Computing Reality: Moore's Law CPU "power" ~2X/18-24 months, constant cost
Genomics Reality Information now ~2X/6 months (Genbank), 4/01
12.4B ~11%/mo
Interoperability; Post-Genomics is the BIG PROBLEM
Biopharma/Economic Reality NCE's/year dropping; R&D increasing
Academic Reality My favorite URLs
Run my algorithms 24x7 !
3October , 2001
Computing Reality: The Net Effect: Take it to the nth
1980 1990 2000
1,000,000X
100,000X10,000
X1,000X
100X
10X
1X
CPU Density/Power, Connectivity/Bandwidth, Node/Value
Moore’sLaw
Gilder’sLaw
Metcalfe’sLaw
Net Effect
4October , 2001
Genomics Reality: Without IT, Data is Just....Data!
Information
Knowledge
Action
Data
5October , 2001
Industry Expectations
7% Market Growth
3-5 NCEs/annum
R&D Costs $350M-$500M
Average sales $265M/annum/drug
Realities
Only realizable in areas of therapeutic and geographical strength
Currently 0.6 NCE/annum
Best estimates are approx $700M/drug
Only 10% achieve >$180M in annual sales
Economic Reality: Drug Discovery is Frustrating!
6October , 2001
BLASTx, FASTA, SMITH-WATERMAN
Phred/Phrap/Consed, Cross_Match, LASSAP
HMMx, CLUSTALx, FrameMatch, D2, NCBI Toolkit
EMBOSS, Artemis, Phylip, Darwin, MAGPIE
BioSCOUT, SRS, BIOPENDIUM, GCG
DoubleTwist, InforMax, ExPASy, ISYS
Oracle, SQL*GT/LIMS, ... Etc.!
x=multiple versions
Key Bioinformatics Software on Solaris
7October , 2001
Academic
U MN, U WI, SDSC/NPACI, NCGR, Wash U (St Louis), Harvard, Rockefeller, CBR-RBC (Canada), UCL, Cambridge U, Humboldt U (Ger), Sydney U, U Queensland, NHRI (Taiwan), Weizmann, InfoBiogen, U. Tokyo, Virginia Bioinformatics Institute, Delaware Bioinformatics Institute, Beijing Genomics Institute, etc.
Commercial
Most BioPharma, Monsanto, Genset, Gene-IT, Keygene, Incyte, OGS, MGW Biotech, DNA Print Genomics, etc.
Selected Major Accounts
8October , 2001
Executive Support for Life Sciences
Dr Greg Papadopoulos Sr
VP & CTO, 2/01
"Sun is committed to working with the life science community to identify and tackle computing/informatics challenges and requirements in the post-genomics era."
9October , 2001
GeoscienceGIS
Weather/ClimateSeismic
EngineeringMCAE
EEeEngineering
Vertical Initiatives
BioX/Comp. Bio.Bioinformatics
ProteomicsPgx
Technology Desktop(Scientific Desktop, Visualization, DCC, Thin Clients and
Development of solution stacks)
Grid Computing(Showcase Implementations of iPlanet-Portal, Sun Grid Engine,
SMC, Sun Clustertools for Tier 1 – Tier 3 Grid Computing)
S&E: Computational Biology Initiative
10October , 2001
Sun Community Support HPC:Grid
HPC Consortium Computational Biology Special Interest
Group COE Informatics Advisory Council (SDN) Events
11October , 2001
Steering Board for every Section
Regional/Global Events
COEs around the world
Industry Collaborations
Compa
ny 2
Center A
Center B
Center C
Center D
Center E
Center FBioComputing
Section
Network of Excellence Centers
Compa
ny 1
12October , 2001
U. Wisconsin – Madison Virginia Bioinformatics Institute Beijing Genomics Institute Delaware Biotechnology Institute ...
Other COEs with CB Components Ohio Supercomputer Center/Children's
Hosp. Cincinnati
COE in Computational Biology
13October , 2001
Summary: Sun's Grid Computing Offerings Sun's existing scalable Grid Computing software stack
Open source building blocks (SGE, Broker, ClusterTools, TCP Portal, Jxta, ...)
Encourage Your research contribution to open source (community)
Integration with Globus etc. (SGE/Broker-Globus-SGEs)
Sun Center of Excellence Program (cooperation ! )
Collaboration, joint Grid projects, Sun GridSIG, . . .
14October , 2001
Different Levels of Grids
Stage 1 - 1 Owner / 1 ClusterCluster Grid Domain of SGE & Technical Computing Stack
Stage 2 - Multiple Owners, 1+ Clusters, 1 Enterprise, 1 SiteCampus Grid Domain of SGE/EE & Multicluster Solutions
Stage 3 - Multiple Sites, Multiple EnterprisesGlobal Grid Domain of SGE/EE plus Grid Frameworks
15October , 2001
Grid Levels
Global Grid (mult.owners, mult.sites)Grid Resource Mgmt, Security, Authentication, Distributed Data
User interface
GL
OB
AL
CA
MP
US
LO
CA
L
Campus 1mult.owners, 1site
Campus 2mult.owners, 1site
Resource Sharing & Brokerage Resource Sharing & Brokerage
Cluster 11owner, 1site
Cluster MgmtResource Mgt
Cluster 21owner, 1site
Cluster MgmtResource Mgt
Cluster 11owner, 1site
Cluster MgmtResource Mgt
Cluster 21owner, 1site
Cluster MgmtResource Mgt
16October , 2001
Grid Computing @ Sun
"The Network is the Computer"
Java, Jini, Jxta, . . .
July'00: Acquisition of Gridware
"Grid" projects since 1995, Julius, Medusa, Eroppa, Unicore, Autobench, . . .
Grid Engine, free, open source, ubiquitous, open API
Department for Grid Computing (inSun VSP): Cluster SW/Stack, Grid SW/Stack, Grid Computing Lab, customer pilots
Sun Grid Computing Council
17October , 2001
Sun Grid Software Stack
Global GridSun TCP, SGE Broker, Globus, Avaki, Cactus, Punch,...
Sun TCP Technical Computing Portal & iPlanet Security
GL
OB
AL
CA
MP
US
LO
CA
L
Campus 1 Campus 2
Sun TCP, SGE Broker Sun TCP, SGE Broker
Cluster 1 TCP SGE ClusterToolsSRM SunMCJxta Jiro QFS
Cluster 2 TCP SGE ClusterToolsSRM SunMCJxta Jiro QFS
Cluster 1 TCP SGE ClusterToolsSRM SunMCJxta Jiro QFS
Cluster 2 TCP SGE ClusterToolsSRM SunMCJxta Jiro QFS
18October , 2001
Sun Technical Computing Portal "prototype"
The only (soon) commercially available hw/sw solution that... Enables quick deployment of tech apps over
Internet, similar to mail and calendaring Combines light-weight architecture with Industry-proven security and system
management Based on iPlanet and Sun Grid Engine
20October , 2001
Sun Distributed Resource Management Load balancing maximizes resource utilization
Transparent job submission & machine selection
Monitoring and accounting ==> SGE Sun Grid Engine, open source
Guaranteeing required resources
Full control over resource utilization
Fair and share based resource usage
Implementation of management policies ==> Sun Grid Engine Broker, open source
21October , 2001
Managing Compute Resources with Sun Grid Engine Broker
Department 1Department 2
Department 3
Department resource accessCampus wide resource demand
Project A
Team B-4
Contractor X
Project C
User 1
Manage the fullmatrix of demand - Users -Teams -Projects
User 2Department 4
Department 5
22October , 2001
Sun Grid Engine Status Ubiquitous, free, open source, open APIs
Current Release: SGE 5.2.3 (July 2001)
Over 12,000 downloads (Sept 2001) (1 Mio downloads in 2084) =>> SGE = The Leading RMS
OpenSource (July 2001) 500,000 lines, Sept 2001: 1000 downloads
Today: Grid Computing everywhere in Sun !! www.sun.com/gridware .../hpc .../edu /...
23October , 2001
Grid Resource Management
LOOSE STRINGENT
SGE BrokerShared Policy Model
Grid Infrastucture Layer (e.g. Globus, Avaki)
SGE Resource Broker
SGEsite C
local policy,user, etc. mgmt.
Cluster Grid
Campus Grid
Global Grid
SGEsite A
local policy,user, etc. mgmt.
SGEsite B
local policy,user, etc. mgmt.
24October , 2001
Demo'd at Argonne National Lab ANL, ARL Army Research Lab, Raytheon, and San Diego SDSC
On 2 SGE clusters (eg: SDSC, 30 cpus and 70 cpus)
Globus/SGE interaction through GRAM scripts
Globus jobs from ANL submitted to ARL cluster
Next step: SGE/EE on top of Globus
Globus & SGE/Broker on Top of SGE
25October , 2001
SGE: Scheduling decisions to select remote site
SGE acting as the resource broker for Globus
Globus: multi-site communication, authentication, security, file transfers,...
SGE/Globus interface to be developed:
SGE/Broker submits and tracks jobs to remote systems using Globus services
SGE/Broker as Part of Globus
26October , 2001
Sun and Open Grid Standards
Example: DRMAA Distributed Resource Management Application API
"The Glue" between Distributed Resource Management and Applications/Tools
=> Makes resource management transparent
Proposed new Working Group at Global Grid Forum in Frascati/Rome, October 2001
Presented by Veridian, Intel, and Sun
27October , 2001
Sun's Grid Strategy Strong Grid core team for developing and productizing
core components (like SGE, Grid Broker,TCP)
Sun Grid Computing Council: Integrate Sun technologies and products and port the environment to all Sun platforms
Sun's partners take care of other computing platforms
Collaborate with the Grid community, IT partners and our customers to build all kinds of different Grids
Sun currently is proposing, designing and building some 50 Grids with research labs, universities and industry Sun Grid software stack available TODAY