An “HPC Cluster Challenge” app for public outreach

Project reference: 1509

EPCC is developing an HPC Cluster Challenge’ web app for phones, tablets and PCs, in which participants will be able to design their own Supercomputer cluster, by picking and assembling predetermined component parts, all while working within a fixed virtual budget. The aim would be to achieve the highest performance score on LINPACK, a well-known benchmark that is used to rank the largest computers in the world! As well as controlling costs, the user must also try to minimise their virtual machine’s power usage-this is one of the key limiting factors for modern-day HPC. The app will explain what the main components of a Supercomputer are, and guide a user through the process of configuring their cluster. Links to more in-depth educational material will also be made available through the app.

Our vision is that the app could be taken to science festivals where we would run a ‘Take The Cluster Challenge!’ event-providing several tablets for visitors to use, and keep a scoreboard on the day. During the SoHPC 2013 and 2014 programmes, students worked with us on the dinosaur simulation racing application, which has proven immensely popular at the large number of HPC outreach events that we do each year. This has shown that a competitive element drives participation and we are aiming to build on this success by complementing it with this educational, HPC focused demonstration. One of the questions we get time and time again from the public is whether they can download and run the dinosaur simulation at home-which, due to the fact that it requires a large parallel machine to perform the work is unfortunately not possible. Therefore an important aspect of this will be to make the cluster design application freely available to download, which will also help widen our outreach target audience to those who might not be able to attend a science festival.

A prototype version will be available by the start of the Summer of HPC project. The main tasks for the student would be to develop the user interface of the app to the state where it is suitable for the general public to use without needing any supervision. This would include design and development of graphics, on line help, and linked educational material. The student would also lead the beta-testing of the app, and fine-tune the cluster model parameters to offer a good ‘gaming’ experience.

 

 

Project mentor: Nick Brown & Amy Krause
Site Co-ordinator: Catherine Inglis

 

Learning Outcomes

The student will learn in detail about the hardware and software elements which make up and HPC cluster and how they affect application performance. They will also develop skills in mobile/web app design, user interface design and gamification.

 

Student Prerequisites (compulsory)

A strong programming background, with an interest in both HPC and the visualisation aspects of the field as well as a flair for design.

 

Student Prerequisites (desirable)

Experience in Java, HTML5, javascript or jquery and/or the willingness to learn these technologies.

 

Training Materials

none

 

Workplan (tentative)

• Work package 1 : (1 week)–SoHPC training week
• Work package 2 : (2 weeks)–Design and implement representative graphics for the various pieces of hardware that can be configured in the app e.g. nodes, CPUs, memory, network, disk, GPUs. Submit workplan.
• Work package 3 : (3 weeks)–Developing supporting material, including help/documentation, and supporting educational material to be made available in-app.
• Work package 4 : (2 weeks)–Tune the model parameters to give a good “gaming” experience, and beta-test the app e.g. with undergraduate students, producing a video demonstration of the app being used. Submit final report.

Final Product Description

The final product will be a mobile/web app, specifically designed to be used at science fairs, schools, and other outreach events. It will be made freely available and could be used as part of PRACE outreach events.

 

Adapting the Project – Increasing the Difficulty

A good student could consider how our cluster performance model could be extended for example to include additional benchmark scores(Streams, PDGEMM, from the HPCC suite), or real applications based on application performance modelling.

 

Adapting the Project – Decreasing the Difficulty

To decrease the scope of the project, we would focus more on the user interface work, and omit the development of the educational material for the future.

 

Resources

All required resources will be provided by EPCC, including suitable development environment and deployment/testing platforms (iOS & Android devices).

 

Organization