DockerCon 2016

1. Using Docker for GPU-accelerated Applications by Felix Abecassis

In addition of being used for visualization, the highly parallel architecture of GPUs also make them a natural fit for accelerating data-parallel and throughput oriented computations such as machine learning or numerical simulations. When GPUs applications are deployed inside data centers they suffer from the same packaging issues as CPU applications, aggravated by a strong need to get reproducible performance results.
The Docker ecosystem is mostly CPU-centric and aims to be hardware-agnostic. This is not the case for GPUs applications since specialized hardware and a specific kernel device driver are now required. We will show how we reconciled those seemingly opposed requirements to enable containerization and execution of GPU applications with Docker.

2. Using Containers and HPC to Solve the Mysteries of the Universe by Deborah Bard

Container technology is being used to answer some of the biggest questions in science today - what is the Universe made of? How has it evolved over time? Scientists use vast quantities of data to study these questions, and analyzing this data requires Big Data solutions on high performance computing resources. In this talk we discuss why containers are being deployed on the Cori supercomputer at NERSC (the National Energy Research Scientific Computing center) to answer fundamental scientific questions. We will give examples of the use of Docker in simulating complex physical processes and analyzing experimental data in fields as diverse as particle physics, cosmology, astronomy, genomics and material science. We will demonstrate how container technology is being used to facilitate access to scientific computing resources by scientists from around the globe. Finally, we will discuss how container technology has the potential to revolutionize scientific publishing, and could solve the problem of scientific reproducibility.

3. Understanding Containers through Gaming by Brendan Fosberry 

Programming games and competitions can be a great way to introduce software development, and motivate people to hone their skills. During the third Global Docker Hackday, we prototyped a simple game platform called “Docker Than Light”. Our goal was to create a fun and competitive exercise to help introduce people to the concept of stateful microservices in containers and gain familiarity with the Docker toolset, all packaged as a competitive “Faster Than Light” style free-for-all.

In this talk we’ll discuss a more unusual implementation of containers; allowing game participants to utilize any programming language to build artificially intelligent actors in a distributed simulation. Docker Swarm and network plugins help distribute resources, maintain isolation and provide a realistic analogy for the different components in the system. We’ll talk about the design and orchestration of the simulation, as well as how the various actor languages and platforms were allowed to interact in a homogeneous way.

4. Tyrion Cannister Neural Styles by Dora Korpar and Siphan Bou

Understanding deep learning is a real challenge, and even getting started installing software on your machine is difficult. In creating our Docker "hack", our goal was to try to make the deep learning algorithm Neural Style accessible to everyone by creating a user-friendly GUI that can be launched with one command and that optimizes the entire experience.