HPC Heterogeneous Cluster

We are going to create an HPC Cluster and would deploy it through Docker. The advantage we get from Docker is that we can work on ?n? number of computer systems simultaneously, increasing the computing power at runtime, and would be independent of whichever Linux distribution we use to deploy t

Project Title

Project Area of Specialization

Information & Communication Technology

Project Summary

We are going to create an HPC Cluster and would deploy it through Docker. The advantage we get from Docker is that we can work on ‘n’ number of computer systems simultaneously, increasing the computing power at runtime, and would be independent of whichever Linux distribution we use to deploy the cluster on; therefore, the concept of heterogeneity as well as elastic computing would be achieved.

Project Objectives

Creating an HPC Heterogenous Cluster environment at runtime, using Docker, and executing a compute-intensive task.

Project Implementation Method

Creating an HPC Cluster, deployed on virtual machine nodes, for conceptual purpose. Generating same HPC Clusters on physical machines and performing benchmark testing on them. Afterwards, implementing MPI libraries on Dockers and deploying them. Then through using containers overlay-network, connecting numerous hosts and deploying docker image on each, hence increasing the computing power at runtime. Execute compute-intensive task on specific HPC Clusters.

Benefits of the Project

To create a cluster at real-time at any given number of infrastructures in minimal time.

The benefit would be, to increase computing power at runtime. Discarding the need of expensive supercomputers, and performing power and time-consuming task though the basic equipment.

The final product would be able to accomplish these tasks by certain means.

-Scientific calculation purpose

-Crypto currency mining

-Video or graphical animation rendering

Technical Details of Final Deliverable

The interface would be any linux distribution running on run-level 3. No graphical user interface at that matter. The user woul require to feed in data manually according to the computation that needs to be executed and let the system take care of the rest. The product is integrated with MATLAB, so all the algorithm and libraries would be in coherence with it.