Preemptive Resource Provisioning for Container-Based Audio/Video Encrypted Collaboration Applications

Abstract

The massive industrial adoption of cloud technology has led to research into cloud-enabling traditional applications. The EMD research project proposes an elastic, reliable, and secure cloud-enabled Audio and Video (A/V) collaboration platform in replacement of a reliable hardware appliance based which had fixed constraints in terms of scalability. In this context, this article introduces heuristics and architectures that efficiently and preemptively allocate EMD’s A/V encrypted and container-based software components in the cloud. A software solution based on Kubernetes, a production-grade container orchestration platform, is compared with another solution focused on dedicated VMs. Both implement resource allocation heuristics that take into account the project’s requirements and location-aware encryption enforcement necessities: encryption is enforced for more sensitive data. A company training scenario with dynamically distributed instructors is modelled using existing A/V stream concepts, and component prototypes are extended to support encryption and containerisation, whose prototype performance evaluation drives the investigation of heuristics and architectures and feeds their larger-scale simulation-based assessment. Results show that container orchestration costs are at least 52% lower than dedicated VMs for this scenario, but rely on relaxing a project requirement: the time taken to establish a new streaming session was to be kept below 2 s. The switch to orchestrated containers raised this up to a maximum of 2.5 s.