36
Santosh Pandey Ram Sharan Chaulagain Prakash Gyawali SUPER COMPUTING USING CLUSTER NETWORK - A supercomputer( HYPE -2 ) Supervisor Prof. Dr. Subarna Shakya

Super COMPUTING Journal

  • Upload
    pandeyg

  • View
    21

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Super COMPUTING Journal

Santosh PandeyRam Sharan Chaulagain

Prakash Gyawali

SUPER COMPUTING USING CLUSTER

NETWORK- A supercomputer( HYPE -2 )

SupervisorProf. Dr. Subarna Shakya

Page 2: Super COMPUTING Journal

“HIGH PERFORMANCE COMPUTING

IS THE NEED OF

FUTURE TECHNOLOGY”

Page 3: Super COMPUTING Journal

THIS IS JUST THE BEGINNING

Page 4: Super COMPUTING Journal

MOORE’S LAW

Page 5: Super COMPUTING Journal

OUR OPTIONS:

MULTIPROCESSOR SYSTEM MULTICOMPUTER SYSTEM

Page 6: Super COMPUTING Journal

PARALLEL PROCESSING

Page 7: Super COMPUTING Journal

Speedup in multiprocessingDepends on parallelizable code

AMDAHL'S LAW

S(P)=Speedup on P processors

T(1)=Time to process in 1 processors

T(P)= Time to process in processors

f=Inherently sequential codep= Parallelizable code

Page 8: Super COMPUTING Journal

High performance computing for researchAchieving super computing at a cheaper rate

than mainframes

OBJECTIVES

Page 9: Super COMPUTING Journal

Muni Sakhya (1980’s)16 nodesFirst and the only one

History of Supercomputer in Nepal

Page 10: Super COMPUTING Journal

Not An Application But Self designed High Performance Computing Architecure And

Cluster Network

Page 11: Super COMPUTING Journal

MiddlewareNetwork ArchitectureMulticore Computers

Developing SuperComputing Framework

Page 12: Super COMPUTING Journal

SYSTEM DESIGN

Page 13: Super COMPUTING Journal

SIMD (Single Instruction Multiple Data)MIMD (Multiple Instruction Multiple Data)

Parallelism Supported By Architecture

Page 14: Super COMPUTING Journal

Every application don’t have same parallelism model

Specific Applications must be programmedExtend Methods of our Architecture

Application Development

Page 15: Super COMPUTING Journal

SYSTEM ARCHITECTURE

Page 16: Super COMPUTING Journal

Computational Model

Page 17: Super COMPUTING Journal

Dynamic Worker Addition and ReductionFault Tolerant Scalable System

FEATURES OF OUR ARCHITECTURE

Page 18: Super COMPUTING Journal

Platform Independent

Page 19: Super COMPUTING Journal

Network TopologyStar Topology

Page 20: Super COMPUTING Journal

Parallel Working

METHODOLGY

Page 21: Super COMPUTING Journal

Server Thread for each Worker at Server side

New Process for each Worker at Client side

Multitasking Model

SERVER

THREAD1

• Provide Chunk 1 To Client1

THREAD 2

• Provide Chunk 2 To Client2

THREAD N

• Provide Chunk N To Client N

Connect to

server

Take chunk to process

ProcessProvide

output to server

Connect to

server

Take chunk to process

ProcessProvide

output to server

Connect to

server

Take chunk to process

ProcessProvide

output to server

Page 22: Super COMPUTING Journal

Running thousands of flops operationsIntegration for finding the value of Pi

Testing

Page 23: Super COMPUTING Journal

RESULT AND ANALYSIS

Page 24: Super COMPUTING Journal

RESULT AND ANALYSIS

1,41,3

2,13,1

3,23,3

4,3

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

Time (ms)

Page 25: Super COMPUTING Journal

RESULT AND ANALYSIS(contd..)

1(10000) 2(20000) 5(50000) 7(70000) 15(150000)0

2

4

6

8

10

12

14

16

18

20

11.954

4.85.9

18

Speedup for 100 Million Iterations

No. of Nodes

Spee

dup

Fig : Exponential Speedup

Page 26: Super COMPUTING Journal

RESULT AND ANALYSIS(contd..)

1(10000) 2(20000) 5(50000) 7(70000) 15(150000)0

2

4

6

8

10

12

14

16

18

20

11.954

4.85.9

18

Speedup for 100 Million Iterations

No. of Nodes

Spee

dup

Theory vs. Practical Data

Page 27: Super COMPUTING Journal

No official data for comparing

Probably the fastest in Nepal

Comparing in Nepal

Page 28: Super COMPUTING Journal

CryptographyData MiningWeather ForecastingResearchArtificial Intelligence

APPLICATION

Page 29: Super COMPUTING Journal
Page 30: Super COMPUTING Journal
Page 31: Super COMPUTING Journal
Page 32: Super COMPUTING Journal
Page 33: Super COMPUTING Journal

Not comparable with bigger super computer due to less nodes

Extension of Architecture library to define new application

LIMITATIONS

Page 34: Super COMPUTING Journal

Supporting Complex ComputationsInter-process Communication for dependent

tasksImplementing GPU for Computation

FUTURE IMPLEMENTATIONS

Page 35: Super COMPUTING Journal

Websites: Don Berker. Robert G. Brown. Greg Lindahl. Forrest Hoffman.

Putchong Uthayopas. Kragen Sitaker. Frequently Asked Questions [Online]. Available: http://www.beowulf.org/overview.faq.html

Technopedia. Computer Cluster [Online]. Available: http://www.technopedia.com/definition/6581/computer-cluster

Dr. Wu-chun. Feng. (2015). The Green500 list- November 2015 [Online]. Available: http://www.green500/list/green201511

Books: Shiflet, Introduction to Computational Science: Modeling and

Simulation for Sciences, Princeton University Press, 2014. Kumar, Lenina, MATLAB: Easy Way to Learning, PHI Learning, 2016. Etter, Introduction to MATLAB, Prentice Hall, 2015 Lemay Laura, Charles L. Perkins, Teach Yourself Java in 21 Days,

Samsnet, 1996.

BIBLIOGRAPHY

Page 36: Super COMPUTING Journal

THANK YOU