Docker vs. KVM: Apache spark application performance and ease of use

In this paper, we compare the performance and ease of use of an Apache Spark application deployed via Docker and alternatively via KVM. Docker containers have rapidly grown in popularity for dynamic cloud computing and for general-use application deployment and collaboration. Recent work comparing the performance of Docker versus VM deployments has found that Docker's reduced overhead results in better performance. We compare Docker versus VM deployments of an Apache Spark application in regard to ease of use in a collaborative development environment. We expect that developing the application in a containerized environment will result in fewer post-deployment tests that fail compared to application versions that are developed in non-virtual or VM environments. Future work will closely examine the performance of these two implementations and more generally investigate potential trade-offs between performance and ease of use.