Building the Teragrid

1
Building the Teragrid

• NPACI All-Hands Meeting
• March 7, 2002
• Carl Kesselman
• Chief Software Architect
• NPACI
• carl_at_isi.edu

2
What are we building?

• A big computer yes
• That would be the tera part
• A fast network.
• Good for getting data in and out
• But what about the grid bit?
• Why are we bothering?

3
Computational Science is a Team Sport

• Collaborations within projects
• PI, co-investigators, students
• National and international collaborations
• Southern California Earthquake Center (SCEC)
• NPACI !!
• Collaborations across disciplines
• Collaboration across resources
• E.g. data federation (National Virtual
  Observatory, BIRN)

4
TeraGrid Application Exemplars

• Traditional supercomputing made simpler
• remote access to data archives and computers
• Distributed data archive access and correlation
• Remote rendering and visualization
• Remote sensor and instrument coupling

5
Teragrid Means New Science
Stimulus Generation
Understanding The Functional Organization of the
Cricket Sensory System (Gwen Jacobs)
Data Acquisition
Grid Middleware
Brain Simulation Signal Analysis
6
The Grid Problem

• Resource sharing coordinated problem solving
  in dynamic, multi-institutional virtual
  organizations

7
Putting the Grid in TeraGrid

• Grid software defines DTF as a system
• Heterogeneity and distribution should be and
  advantage!
• Grid infrastructure
• Intersite security, scheduling, discovery,
  resource coordination, resource management,
• TeraGrid poses unique challenges
• Issues of scale and performance
• Work at NPACI (and Alliance) has positioned us
  well to address these problems

8
Layered Grid Architecture(By Analogy to Internet
Architecture)
Application
9
Teragrid is not an Island

• Four DTF sites a core
• Must form an integrated facility
• End user sites important components
• Strong need to integrate into other high-end and
  specialized facilities
• Experimental facilities, PSC, HEP Tier II sites,
  campus facilities

10
NSF Middleware Initiative

• NSF Funded Project to build national middleware
  infrastructure
• USC/ISI, SDSC, U. Wisc., ANL, NCSA, I2
• Software Integration (NMI Software Releases)
• Interoperability
• Testing
• Install, Configure, Manage
• University Campus Infrastructure Integration
• Campus Authentication / GSI
• Enterprise Directories / GSI and MDS
• Use NMI as Teragrid Baseline
• Specialize for Teragrid unique aspects (e.g. Viz
  resources)

11
NMI-R1 Software Components

• Globus Toolkit
• Condor-G
• Network Weather Service
• KX.509 / KCA
• Certificate Profile Maker
• Pubcookie
• Grid Packaging Tools

12
NPACI Specific Teragrid software

• Applications via alpha projects
• Already layered on NPACI technologies
• Need to refocus NPACI Grid technologies for
  TeraGrid
• Build on NMI foundation
• Contribute to NMI foundation

13
Legion-G Architecture
Grid-enabled DFS
CHARMM
LegionFS
Tools for Easily Creating and Invoking Grid
Services
Legion GUIs for P-Space Studies
CCA
SRB
Legion-G Grid-Enabled Object Model
NWS
Legion-G Object Process Instantiation
Integration with LegionFS
Security in and across Legion-G Objects
Information Services for Legion-G Objects
GSI
GRAM
GRIP
GridFTP
14
Immediate Term Legion-G and the DTF

• Transition existing NPACI Legion use and
  infrastructure to Legion-G
• Persistent infrastructure and tools for
  scientific discovery
• Protein Folding on the Grid CHARMM portal and
  Legion-G support (with Charles Brooks III, TSRI)
• Exposure of PDB via LegionFS
• Work with scientists to use Legion-G for
  collaboration in the DTF

15
DataCutter

• Purpose Specialized components for processing
  data
• Based on Active Disks research Acharya, Uysal,
  Saltz ASPLOS98,
• filters logical unit of computation
• reads data from input buffers, compute/filter/aggr
  egate data, then writes data to output buffers
• filter can only carry out subsetting,
  commutative/associative data aggregations
• subsetting implemented by (among other tings)
  multilevel hierarchical indexes based on R-tree
  indexing method.
• filter computations are pipelined
• streams how filters communicate
• unidirectional buffer pipes
• application developer specifies connectivity
  between filters
• copies because of the way filters are defined,
  each filters computations can be carried out by
  a system defined number of transparent copies

16
Teragrid Data Processing Infrastructure
DataCutter/Globus/NWS

• DataCutter
• filters subset, filter, aggregate data
• Network Weather Service
• used to provide network, processor performance
  information used in placing filter copies,
  determining which data replicas to use
• Globus
• used to run filter code, track location of data
  replicas, maintain updated NWS performance
  information, record location of active filter
  copies.

17
Storage Resource Broker

• Current
• Authentication
• GSI certificate mapped to local user
• local user account authenticated to SRB
• Access
• GSI-FTP callbacks mapped to SRB client calls
• Future
• Authentication
• Pass GSI certificate directly to SRB
• Access
• GSI-FTP callbacks mapped to SRB server calls.
• Support for creating/specifying containers,
  replication,subsetting

18
NWS as an Grid Information Provider
GIIS
GRIP
GRRP
GIIS
GRIS
sensor data
19
Its Also a Client
NWS
Sensors
NWS Process Registrations
Forecasters
GIIS or GRIS
Reporters
Sensor Data
Persistent State
20
Summary

• Teragrid is an opportunity for broadening the
  scope of computational science
• Infrastructure must enable applications
• Teragrid is about a system
• Grid middleware a critical part of the
  infrastructure
• Integration into the global Grid environment is
  essential
• NPACI technology can
• Provide value added
• Provide application transition
• Lot of work in the next three years
• Logistics and development