Design development and bench-marking of HPC for Machine Learning

Project Title

Design development and bench-marking of HPC for Machine Learning

Project Area of Specialization Cloud Infrastructure,Project Summary

The aim of my project is to develop a high performance computing Beowulf cluster to increase the processing power of the available hardware and to benchmark the performance by running simulations for machine learning on the developed unit.

Project Objectives

(a) Design and development of a cluster of computers.
(b) Selection,installation and configuration of software for parallel execution on the designed cluster.
(c) Benchmarking the performance of software for machine learning software on the developed cluster.

Project Implementation Method

For a beawulf cluster to function we need a master and a compute node. In my case I will be using 3 compute nodes and 1 master node but master node will be used in calculation of simulation. Four steps are involced

1. A network connection between the master and the compute nodes

2. Shared file directory amoung the master and compute nodes

3. OpenMPI for communication amoung nodes

4. The software which going to be parallesed 

Benefits of the Project

Beowulf Clusters have the best performance/cost ratio compared to standalone supercomputers. The overall advantages are scalability, performance/cost, flexible configuration/upgrade, able to keep up with change in technology, higher level of control for users.  These advantages make Beowulf Clusters capable of being used for many applications that a standalone supercomputer would be too expensive for and tasks that standalone supercomputers are good at. In short this project will enable the people 

Technical Details of Final Deliverable

(a) A thorough analysis of the software for Machine Learning Software. analysis in term of processing, networking, storage and memory requirement.
(b) Demonstration of a scalable architecture of the high performance computing structure based on an appropriate individual workstation.
(c) Demonstration of a whole system scalable software architecture consisting of an appropriate selection of OS and middleware.

Final Deliverable of the Project HW/SW integrated systemType of Industry Education , Others Technologies Artificial Intelligence(AI), Big DataSustainable Development Goals Reduced InequalityRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	69000
Processor	Equipment	4	6000	24000
Network Switch	Equipment	1	1000	1000
Motherboard	Equipment	4	7000	28000
RAM	Equipment	4	1500	6000
Hard Disk	Equipment	4	2500	10000