Utility Aware Placement of HPC Workloads in Heterogeneous Infrastructure Cloud

11th April 2017

surya

Dr. Surya Narayanan Natarajan

High Performance Computing Researcher


11:30  -  12:00
Q303, DCU Business School

In the IaaS cloud business model, there are 2 communities:
1. Customer and
2. Infrastructure provider.

Customers seek good performance and user experience while providers seek to optimise profitability, where customer satisfaction, optimizing utilization of resources and total cost of service delivery play significant roles. There are various attributes that contribute to customer and provider objectives. In the future autonomous data center, a resource manager along with analytical capabilities and job scheduler can be tasked to handle this trade-off and make a decision on how to schedule and place a job in the available resource such that both the parties can be satisfied. There may be various influencing parameters or ‘attributes’ that needs to be considered in making the placement decisions that can be modelled using utility theory, and ultimately, a placement decision can be taken by looking at the utility values depending on the business needs. We use OpenFOAM, a CFD application as a use case to demonstrate an approach that uses utility based analytics in finding better placement solution in the heterogeneous infrastructure cloud.


Session Category :  Session 4 | HPC in The Cloud Use Cases