Grid Computing Research at IITM - PowerPoint PPT Presentation

1 / 21
About This Presentation
Title:

Grid Computing Research at IITM

Description:

Distributed systems, parallel computing on clusters, ... SSA: Sequential Algo. TMA: Temperature Modifier Algo. LE : Locking Edges Algo. MLE: Modified LE ... – PowerPoint PPT presentation

Number of Views:423
Avg rating:3.0/5.0
Slides: 22
Provided by: lic76
Category:

less

Transcript and Presenter's Notes

Title: Grid Computing Research at IITM


1
Grid Computing Research at IITM
  • D Janaki Ram, Professor
  • Distributed Object Systems Lab,
  • Professor, Dept. of CS E
  • Indian Institute of Technology, Madras

2
Key Focus
  • Engineering Software Systems
  • Enabling technology for next generation
    applications
  • Distributed systems, parallel computing on
    clusters, software engineering, databases, object
    orientation and CAD/CAM
  • Modeling and system building

3
Cluster Computing Initiatives of DOS Lab
  • Distributed Problem Solving
  • Anonymous Remote Computing (ARC)
  • Anonymous Remote Computation and Communication
    (ARCC)
  • P-CORBA Middleware based cluster computing
  • Moset An Anonymous Remote Mobile Cluster
    Computing Paradigm
  • SnehaSamuham ARCC over the Internet

4
Distributed Problem Solving
  • Genetic Clustering Algorithm (GCA)
  • Optimization problem
  • Based on crossover
  • Implemented on Distributed Problem Solver (DiPS)
  • GCA for Job Shop Scheduling (JSS)
  • GCA for Travelling Salesman Problem (TSP)

5
Implementation Results
SSA Sequential Algo TMA Temperature Modifier
Algo LE Locking Edges Algo MLE Modified LE
C Cost of best solution obtained T Execution
time (secs) N No. of cooperating nodes Problem
Size Number of jobs Number of machines
6
Performance of CA and CGA for a specific
ProblemInstance (Size 1015)
7
Anonymous Remote Computing (ARC)
  • Model for parallel programming over loosely
    coupled distributed systems.
  • Co-existing sequential and parallel loads
  • Parallel program collection of Remote
    Instruction Blocks (RIBs)
  • Dynamic load balancing migration of RIBs
  • Horse power factor abstraction of computation
    power of a node

8
p1
p1
p1
p2
p2
p3
p1
p3
completed RIB
executing RIB
Growing and Shrinking Remote Instruction Blocks
(RIBs)
9
The Heterogeneity Load Test, 16 Processors
10
ARCC
  • Transparent programmability of communicating
    parallel tasks in a cluster
  • Distributed pipes location independent data flow
    between tasks
  • Programs
  • resilient to changing conditions of Network of
    Workstations (NOWs)
  • Devoid of network related code
  • Overlap communication and computation
  • Nullify communication overhead

11
ARCC
Solaris Node
HP-UX Node
Linux Node
High HPF
Low HPF
Max HPF
12
DP Results
13
P-CORBA Transmission and Distribution of
Computing Power over CORBA Bus
  • Notion of concurrency into CORBA
  • Object migration service
  • Message filter client transparent request
    redirection
  • Load balancing service
  • Dynamic, fully decentralized, minimal overhead
  • Communicating tasks XML for data transfer

14
Speed up for Steady State Problem P-CORBA
15
Sneha-Samuham Computational Model Pictorial View
Parallel Task
16
Advantages of Sneha-Samuham over MPI.
17
Moset An Anonymous Remote Mobile Cluster
Computing Paradigm
Parallel Applications
Parallel Applications
Sequential Applications
Parallel Applications
Sequential Applications
Sequential Applications
Parallel Programming API
Moset Computing Kernel
Fault-Tolerant Exactly-once Reliable Multicast
Protocol
Cluster Interconnection - Wired
Network Wireless Network
MH
MH
MH
MH
18
Moset Results
19
Real World Problems Solved
  • Parallel simulated annealing
  • Iterative grid computation
  • Steady state problem computational fluid
    dynamics
  • Navier stokes solver
  • Parallel image rendering
  • Neutron shielding simulation

20
Current Research in the Area
  • Grid computing
  • Extending campus wide grid into a worldwide grid
  • Mobile Grid
  • Middleware for dynamic service composition for
    integrating mobile devices into the grid
  • Integration of data and computing grids

21
Key Publications
  • Rushikesh K. Joshi and D. Janaki Ram, Anonymous
    Remote Computing A Paradigm for Parallel
    Programming on Interconnected Workstations, IEEE
    Transactions on Software Engineering, Vol.25,
    No.1, Jan, 1999, pp. 75-90.
  • Binu K.J., Karthikeyan R. and D. Janaki Ram, DP
    A Paradigm for Anonymous Remote Computation and
    Communication for Cluster Computing, IEEE
    Transactions on Parallel and Distributed Systems,
    Vol. 12, No.10, October 2001, pp. 1-14.
  • D. Janaki Ram, T. H. Sreenivas and Ganapathy
    Subramaniam, Parallel Simulated Annealing
    Algorithms, Journal of Parallel and Distributed
    Computing, Vol.37, No. 2, 1996, pp. 207-212.
  • Binu K. J. and D. Janaki Ram, Integrating Task
    Parallelism in Data Parallel Languages for
    Parallel Programming on NOWs, Concurrency -
    Practice Experience, 200012, pp. 1291-1315.
  • R. K. Joshi and D. Janaki Ram, Object Based
    Subcontracting for parallel programming on
    loosely coupled distributed systems. Journal of
    Programming Languages, 4 (1996) pp. 169-183.
  • D. Janaki Ram and A. Vijay Srinivas, Object
    Migration in CORBA The Journal of The Computer
    Society of India, Vol.32, No.1, March 2002, pp.
    18-27.
  • D. Janaki Ram, A. Vijay Srinivas and P. Manjula
    Rani, A Model for Parallel Programming Over
    CORBA, Journal of Parallel and Distributed
    Computing Vol. 64, No. 11, November 2004, pp.
    1256-1269.
  • M. A. Maluk Mohamed, A. Vijay Srinivas and D.
    Janakiram, Moset An Anonymous Remote Mobile
    Cluster Computing Paradigm, To appear in Special
    Issue on Design and Performance of Networks for
    Super-, Cluster-, and Grid-Computing to appear in
    the Journal of Parallel and Distributed Computing
    (JPDC).
Write a Comment
User Comments (0)
About PowerShow.com