Title: High-performance Cloud Computing for Symbolic Computation

Proposer: Hans-Wolfgang Loidl

Suggested supervisors: Hans-Wolfgang Loidl

Goal: Assess High-performance Cloud Computing for Symbolic Computation

Description:

A current trend in building large-scale computer science and IT frameworks is to use centralised Cloud services and to out-source functionality that was previously provided locally. Several open source frameworks for realising such Cloud services have emerged.

The goal of this project is to use current, open-source Cloud software as the platform for high-performance computing on stock hardware. Using an existing parallel symbolic application, the determinisation of a non-deterministic finite state automaton (developed in the CALCIUM project), performance should be evaluated on our Beowulf cluster, directly using MPI and running it inside virtual machines coordinated by a Cloud infrastructure such as OpenNebula. Raw performance as well as usability and performance portability should be assessed on both platforms.

The milestones in this project will be

• Literature review on frameworks for Cloud computing and on the application domain
• Installing a cloud computing framework and the parallel application
• Comparing the performance on the Beowulf cluster
• Improving the parallel performance on the cloud platform
• Assessment of performance, usability and performance portability

Resources required: Linux machines

Degree of difficulty: Moderate

Background needed: Parallel programming knowledge, general systems skills.

References:

1. Open Nebula
2. CALCIUM