Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

1

GARUDA National Grid Computing Initiative

N. Mohan RamChief Investigator – GARUDA

9th February 2006Kolkatta

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

2

Presentation Outline

• Overview

• Technologies and Research Initiatives

• Communication Fabric

• Resources

• Partners

• Applications

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

3

Project Overview

• Precursor to the National Grid Computing Initiative

– Test Bed for the grid technology/concepts and applications

– Provide inputs for the main grid proposal

• Major Deliverables– Technologies, Architectures, Standards & Research

Initiatives– Nation-wide high-speed communication fabric– Aggregation of Grid Resources– Deployment of Select applications of National Importance– Grid Strategic User Group

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

4

Technologies, Architectures, Standards

and Research Initiatives

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

5

Deliverables

• Technologies– Garuda Component Architecture & Deployment– Access Portal– Problem Solving Environments – Collaborative Environments– Program Development Environments– Management and Monitoring– Middleware and Security– Resource Management and Scheduling– Data Management– Clustering Technologies

• Research Initiatives– Integrated Development Environments– Resource Brokers & Meta Schedulers– Mobile Agent Framework– Semantic Grid Services (MIT Chennai)– Network Simulation

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

6

Garuda Component Architecture

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

7

C-DAC Development & Deployment CollaborationsResearch Initiatives OpenCommercial

Access GRIDAccess GRID Video Conferencing over IPVideo Conferencing over IP

Grid ApplicationsGrid Applications

MDSMDS

MPICH-G2MPICH-G2

Grid SchedulersGrid Schedulers

SUN Grid EngineSUN Grid EngineLoadlevelerLoadleveler

Storage Resource BrokerStorage Resource Broker Visualization SoftwareVisualization Software

PDE

Collaborative Environment

Benchmarks& Applications

Middleware & Security

Storage & Visualization

Monitoring &

Managem

ent

C-D

AC

GridM

onC

-DA

C G

ridMon

NM

SN

MS

Ganglia

Ganglia

DIViA for GridDIViA for Grid IDEIDE CactusCactusWorkflowWorkflow ProfilersProfilers

Grid ProbesGrid Probes

Integration & EngineeringIntegration & Engineering

GLOBUS 2.x/4.x

Certificate AuthorityCertificate Authority

Resource BrokerResource Broker

Grid SecurityGrid Security

Semantic Grid ServicesSemantic Grid Services

C-DAC Grid PortalC-DAC Grid Portal Problem Solving EnvironmentsProblem Solving EnvironmentsGrid AccessMethods

GARUDA Components

C-DAC Development & Deployment CollaborationsResearch Initiatives Open SourceCommercial

Access GRIDAccess GRID Video Conferencing over IPVideo Conferencing over IP

Grid ApplicationsGrid Applications

MDSMDS

MPICH-G2MPICH-G2

Grid SchedulersGrid Schedulers

SUN Grid EngineSUN Grid EngineLoadlevelerLoadleveler

Storage Resource BrokerStorage Resource Broker Visualization SoftwareVisualization Software

PDE

Collaborative Environment

Benchmarks& Applications

Middleware & Security

Storage & Visualization

Monitoring &

Managem

ent

C-D

AC

GridM

onC

-DA

C G

ridMon

NM

SN

MS

Ganglia

Ganglia

Monitoring &

Managem

ent

C-D

AC

GridM

onC

-DA

C G

ridMon

NM

SN

MS

Ganglia

Ganglia

DIViA for GridDIViA for Grid IDEIDE CactusCactusWorkflowWorkflow ProfilersProfilers

Grid ProbesGrid Probes

Integration & EngineeringIntegration & Engineering

GLOBUS 2.x/4.x

Certificate AuthorityCertificate Authority

Resource BrokerResource Broker

Grid SecurityGrid Security

Semantic Grid ServicesSemantic Grid Services

C-DAC Grid PortalC-DAC Grid Portal Problem Solving EnvironmentsProblem Solving EnvironmentsGrid AccessMethods

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

8

Garuda Resource Deployment(at C-DAC centres)

AIXCluster

Shared User SpaceShared Data Space

Linux Cluster

Bangalore

Solaris Cluster

LinuxCluster

Shared User SpaceShared Data Space

Pune

Chennai HyderabadResource

Manager for Grids

Linux Cluster

ResourceManager for Grids

Garuda Access Portal

Other UsersPune Users

End Users access the grid through the Garuda Access Portal

HighAvailability

HighAvailability

Resource Manager for Grids

HighAvailability

HighAvailability

Resource Manager for Grids

Bangalore Users

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

9

Garuda Access Portal

• Addresses the usability challenges of the Grid

• Supports submission of parallel and sequential jobs

• Support for Accounting

• Integration with Grid Scheduler under progress

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

10

Collaborative Environments

• Enable Collaborative environment for the Grid developers, users/partners. Will facilitate development team meetings and collaborative project design/progress reviews

• IP based video conferencing over the high speed communication Fabric

• Initial Target : Enable all centres of C-DAC participating in the Garuda development & deployment to collaborate through video conferencing

• Also exploring Access Grid environment

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

11

Program Development Environment

• Enable users to carry out entire program development life cycle for the Grid

• DIViA for the Grid

• Features– Supports MPICH-G2

debugging– Communication and

computational statistics in different graphical formats

– Identification of potential bottlenecks

– Unique method of tracing, leads to enhanced information with reduced log file size

• Debugger in Design Phase

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

12

Management and Monitoring

• Monitors status & utilization of the Grid components : compute, network, softwares etc.

• Used by System Administrators and end users

• Being deployed at the Grid Monitoring and Management Centre(GMMC)

• User friendly interface

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

13

Middleware & Security

• Deployed using Globus Toolkit, Commercial and C-DAC developed components

– GT2 for operational requirements– GT4 for research projects

• Resource Management and Scheduling– Moab from Cluster Resources for Grid Scheduling– Local Scheduling using Load Leveler for AIX Clusters

and Torque for Solaris and Linux Clusters

• Data Management– Storage Resource Broker from Nirvana for Data Grid

functionalities

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

14

Resource Management and Scheduling

• Grid Scheduler from Cluster Resources– Industry Leading Scheduler– Components include Moab Workload Manager, Moab Grid

Scheduler and Moab Cluster Manager– Integrates with Globus

– Data Management through GASS and GridFTP– Job staging with GRAM/Gatekeeper services– User management through Globus user mapping files– Security through X509-based client authentication

• Grid Scheduler Features– Intelligent Data Staging– Co-Allocation & Multi-Sourcing– Service Monitoring and Management– Sovereignty (Local vs. Central Management Policies)– Virtual Private Cluster and Virtual Private Grid

• Local Resource Managers– Load Leveler on AIX– Torque on Solaris/Linux clusters

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

15

Wide Area Grid

Administrator:Sets policies and manages via Cluster Manager for his or her own cluster and via Grid Resource Manager for

Grid policies

End Users: (In Multiple User Spaces)

Submit jobs via Garuda Grid Access Portal

GlobusGlobus

User Space 1User Space 1 User Space 2User Space 2 User Space NUser Space N

Grid FTPGrid FTP

Grid Resource Manager Interacts with Grid FTP to

Stage Data to each of the Clusters

Grid Resource Manager Leverages

the Security and Access Control

provided in Globus

Gri

d R

eso

urc

e M

anag

erG

rid

Res

ou

rce

Man

ager

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

16

Local Area Grid (C-DAC Bangalore)

Administrators:Sets policies and

manages via Moab Cluster Manager

Moab Cluster ManagerMoab Cluster Manager

Moab Cluster Manager:Acts as the Interface, using

wizards and forms to improve ease of use and to unify interface to

Workload and Resource Managers

Garuda Access PortalGaruda Access Portal

End Users: (In a Single User Space)Submit jobs via web form Interface

Single User SpaceSingle User Space

Moab Workload ManagerMoab Workload Manager

Solaris Cluster Head Node

TorqueTorque

Moab Workload Manager: Enforces policies, monitors

workload and controls submissions through resource manager

Load LevelerLoad Leveler TorqueTorque

Linux Cluster Head Node AIX Cluster Head Node

…..

Unified Data Space

….. …..

OS and CommunicationOS and Communication OS and CommunicationOS and Communication OS and CommunicationOS and Communication

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

17

Data Management• Enable data-oriented applications via an integrated

but distributed storage and data management infrastructure

• Requirements– Heterogeneous Data Access across Multiple Locations– Data Security– Reliability and Consistency of Data– Support for Unified Namespace and Multiple File Systems– Optimal turn-around for Data Access

– Parallel I/O– Bulk Operations– Intelligent Resource Selection and Data Routing– Latency Minimization

– Vertical and Horizontal Scalability

• Garuda Data Grid– Storage Resource Broker from Nirvana

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

18

Clustering Technologies

• Software– High Performance Compilers– Message Passing Libraries– Performance and Debugging Tools– I/O Libraries, Parallel File System– Cluster Management Software– Available for AIX, Solaris and Linux

Clusters

• Hardware– 5Gbps SAN Technologies completed– Reconfigurable Computing Systems

for bioinformatics & cryptanalysis under progress

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

19

Research Initiatives• Resource Broker

– Standards are yet to be formulated– Match the user requirements with the available resources– Address co-allocation of computation and communication– Forecasting the availability of resources

• Grid IDE– Writing and enabling applications to exploit the Grid– Compiling/cross-compiling across different platforms

– Seamless integration of complex functionalities – Support for multiple programming interfaces

• Semantic Grid Services (MIT, Chennai)– Publishing Grid Services– Intelligent discovery of Grid services– Integration with Garuda Portal

• Mobile Agent Framework– Monitoring of resources in the Grid– Grid software deployment and maintenance

• Network Simulation– Inputs for the next phase fabric architecture – To study impact of changes in traffic profile on the performance

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

20

Garuda Communication Fabric

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

21

Objectives & Deliverables

• Objective– Provide an ultra-high speed multi services

communication fabric connecting user organizations across 17 cities in the country

– Provide seamless & high speed access to the compute, data & other resources on the Grid

– In Collaboration with ERNET

• Deliverables– High-speed Communication Fabric connecting 17 cities– Grid Management & Monitoring Centre– IP based Collaborative Environment among select

centres

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

22

Fabric Connectivity

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

23

Features

• Ethernet based High Bandwidth capacity

• Scalable over entire geographic area

• High levels of reliability

• Fault tolerance and redundancy

• Interference resilience

• High security

• Effective Network Management

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

24

Grid Management & Monitoring Centre(GMMC)

• To provide an integrated Grid Resource Management & Monitoring Framework

• Network Traffic Analysis and Congestion Management

• Change and Configuration Management

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

25

Grid Resources

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

26

Objective and Deliverables

• Objective– Provide heterogeneous

resources in the Grid including Compute, Data, Software and Scientific Instruments

– Deploy Test facilitates for Grid related research and development activities

• Deliverables– Grid enablement of C-DAC resources

at Bangalore and Pune

– Aggregation of Partner Resources

– Setting up of PoC Test Bed and Grid Labs at Bangalore, Pune, Hyderabad and Chennai

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

27

Resources

• HPC Clusters & Storage from C-DAC– Bangalore : 128 CPU AIX Cluster,5 TB Storage– Pune : 64 CPU Solaris Cluster

: 16 CPU Linux Cluster, 4 TB Storage– Chennai : 16 CPU Linux Cluster, 2 TB Storage– Hyderabad : 16 CPU Linux Cluster, 2 TB Storage– The proposed 5 TF system to be part of the Grid

• Satellite Terminals from SAC Ahmedabad

• 2 TF Computing Cycles from IGIB Delhi

• 32 way SMP from Univ. of Hyderabad

• 64 CPU cluster from MIT, Chennai

• 64 CPU cluster from PRL, Ahmedabad

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

28

Grid Partners

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

29

Motivation and Status

• Motivation– Setup a User Group to Collaborate on

Research and Engineering of Technologies, Architectures, Standards and Applications in HPC and Grid Computing

– To Contribute to the aggregation of resources in the Grid

• Current Status– 37 research & academic institutions in the

17 cities have agreed in principle to participate

– ERNET-HQ in Delhi– 7 centres of C-DAC– Total of 45 institutions

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

30

Partner Participation Institute of Plasma Research, Ahmedabad Physical Research Laboratory, Ahmedabad Space Applications Centre, Ahmedabad Harish Chandra Research Institute, Allahabad Motilal Nehru National Institute of Technology, Allahabad Jawaharlal Nehru Centre for Advanced Scientific Research, Bangalore Indian Institute of Astrophysics, Bangalore Indian Institute of Science, Bangalore Institute of Microbial Technology, Chandigarh Punjab Engineering College, Chandigarh Madras Institute of Technology, Chennai Indian Institute of Technology, Chennai Institute of Mathematical Sciences, Chennai

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

31

Partner Participation (Contd.) Indian Institute of Technology, Delhi Jawaharlal Nehru University, Delhi Institute for Genomics and Integrative Biology, Delhi Indian Institute of Technology, Guwahati Guwahati University, Guwahati University of Hyderabad, Hyderabad Centre for DNA Fingerprinting and Diagnostics, Hyderabad Jawaharlal Nehru Technological University, Hyderabad Indian Institute of Technology, Kanpur Indian Institute of Technology, Kharagpur Saha Institute of Nuclear Physics, Kolkatta Central Drug Research Institute, Lucknow Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

32

Partner Participation (Contd.) Bhabha Atomic Research Centre, Mumbai Indian Institute of Technology, Mumbai Tata Institute of Fundamental Research, Mumbai IUCCA, Pune National Centre for Radio Astrophysics, Pune National Chemical Laboratory, Pune Pune University, Pune Indian Institute of Technology, Roorkee Regional Cancer Centre, Thiruvananthapuram Vikram Sarabhai Space Centre, Thiruvananthapuram Institute of Technology, Banaras Hindu University, Varanasi

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

33

Applications of Importance for PoC Garuda

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

34

Objectives and Deliverables

• Objectives– Enable applications of

national importance requiring aggregation of geographically distributed resources

• Deliverables– Grid enablement of

illustrative applications and some demonstrations such as

– Bioinformatics– Disaster Management

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

35

Bioinformatics

• Bioinformatics Resources & Applications Facility (BRAF) on PARAM Padma

• Supports highly optimized Bioinformatics codes on the PARAM Padma

• Web computing portal providing all computational facility to solve related problems

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

36

Disaster Management

PARAM Padma

at Bangalore

Flight data transmission from

nearby Airport

Grid Partner Resource

at Pune

User Agencies

GRID Communication

Fabric

High Speed Communicati

on

UserAgenci

es

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

37

Disaster Management (contd..)• Requirements

– Timely dissemination of disaster information to user agencies

– Organize logistics around automated and secure work flow and data base

• Challenges– Widely spread application resources and

types : disaster sensors, compute, application experts– Turn around time for the work flow

Natio

nal G

rid C

om

pu

ting

Initiative - G

arud

a Fe

bru

ary

200

6 C

-DA

C /

M

oh

an

Ram

N

38

Thank you!