Workload optimization through heterogeneous and low power accelerators targeting Cloud computing systems
Thesis Type: Master Thesis for Computer Science, Computer Engineering
• MS students in Electronic Engineering/Computer Science
• Experience with main programming languages (C/C++, Python), basic knowledge of processor architectures.
Cloud Computing and HPC infrastructures are rapidly evolving to embrace a large set of heterogeneous and low power computing devices. Such enormous variety of devices is necessary to improve energy efficiency of modern datacenters. On the other hand, it represents a challenge from the programming perspective. Moreover, this variety of processing elements make difficult to distribute workloads in an efficient manner.
The objective of the work is to start developing applications and benchmark kernels that take advantages from heterogeneous parallel (low power many-core) architectures (for example: FPGA, GPU…). During the work’s activity, the candidate will be focused on the development of parallel version of heuristic algorithms (e.g., genetic algorithms, PSO, ant colony, etc.) targeting the workload optimization of servers. The candidate will use the Parallella platform, a modern development board equipped with a many-core accelerator.
Contact: send CV to firstname.lastname@example.org specifying the thesis code and title.