We are excited to welcome another cohort of graduate students (or exceptional undergrads!) for Chameleon's Summer 2021 internships. Applications are due March 14th, so if you're interested, apply early! All you need is your resume!

The link to the application is here: https://bit.ly/3svQDcW

Project Descriptions: 

Project 1: CHI@Edge: From Edge to Cloud Exploration to Driving Autonomous Cars

The edge to cloud computing is currently in high demand for both research and instructional purposes. With the rise of virtual instruction, there is a growing need for remotely accessible, fun lab exercises that allow students to interact directly with sensors and devices. Chameleon will support this need with infrastructure called CHI@Edge. In order to seed the usage of this new type of resource in the testbed and give insight into how the provided capabilities can/should be used in practice, this project will develop several instructional modules that will develop exercises exploring various aspects of interaction between edge and cloud.

The exercises will range from data acquisition via e.g., providing a simple data capture via Raspberry Pi’s GPIO (General Purpose Input/Output) connector and simple analysis, through measuring the impact of latency/bandwidth of acquired data on overall processing time of an edge to cloud application, to illustrating the use of machine learning algorithms on the edge device using a model trained in the cloud. These simple exercises will be combined in a workflow for development of an autonomous model car, with the controller being a raspberry pi + camera.

The project will explore the use of a range of edge devices from general purpose (Raspberry PIs) to AI/ML class hardware, including NVIDIA’s Jetson TX2, Nano, and Xavier NX as well as potentially different software systems such as EdgeX Foundry. It will involve working with edge-computing software and hardware stacks, interfacing those stacks with cloud technologies, and enabling an edge-to-cloud workflow for experimentation and teaching. The exercises will be developed as Jupyter notebooks, using Chameleon’s Jypyter integration.

 

Project 2: Enabling Edge to Cloud Experiments with Autonomous Vehicles in the Field

The ability to experiment in the Edge to Cloud continuum is currently in high demand. Chameleon supports this need with infrastructure called CHI@Edge which allows users to allocate edge devices on the testbed and deploy edge to cloud experiments. In order to seed the usage of this new type of resource in the testbed and give insight into how the provided capabilities can/should be used in practice, this project will create an experiment exploring various aspects of interaction between edge and cloud.

FIU's Boswell Laboratory for Marine Ecology and Acoustics has built and deployed an Autonomous Surface Vehicle (ASV) platform integrated with several onboard sensors (sonar, acoustic, biological, camera.) In partnership with researchers at FIU, this project will work to prototype and evaluate new configurations that address current science use-cases by exploring various trade-offs between processing data collected by autonomous vehicles at the edge and moving it to the cloud.

This project will explore a range of edge devices, use open-source IoT software platforms such as EdgeX Foundry, and interface with cloud and machine learning technologies to determine how to best leverage existing research cloud infrastructure such as Chameleon to support experiments on the edge.

 

Project 3: Examining Pre-Emptible Instances in Clouds

Using pre-emptible instances in a cloud can increase utilization by filling gaps between allocations with fixed reservation times: the cloud infrastructure can pre-empt these instances if it requires access to those resources by other tasks. Commercial clouds have their implementations and policies on preemptible instances, such as AWS spot instances and Google preemptible VM instances. We would like to provide a similar functionality in Chameleon, a NSF-funded cloud testbed based on uses OpenStack.

The specific tasks will include running high-throughput jobs as preemptible instances on Chameleon, research on policies for preemptible instances on various cloud platforms and testbeds, explore how different policies will affect the utilization and user experience of Chameleon testbed, as well as proposing and implementing a policy for using preemptible instances on Chameleon.

This project will involve modifications to the open source OpenStack platform, training ML modules to reflect different policies, and performance analysis.