Performance Comparison of Virtualization Methods
Anushka Hebbar, Shree Thavarekere, Anusha Kabber, D Shruvi, K. V. Subramaniam · 2022
Cloud service providers can choose from a plethora of virtualization technologies for containerized application deployment, or in the case of serverless providers, running functions from multiple tenants. Generally, a cost-benefit analysis is performed to determine the choice of deployment platform. Containers and traditional virtual machines have, till recent history, been dominant choices on the market. Containers provide quick boot times, high speeds, and efficient resource utilization, whereas virtual machines are a set standard for secure isolation among multiple tenants. Guaranteeing all of the performance requirements - solid security, superior speeds, low latency bootup, and efficient resource utilization - has been impossible due to the relative drawbacks of these platforms. To tackle this constraint, AWS’s Firecracker, a virtual machine monitor (VMM), was created to provide the best of all worlds, by claiming to combine the security/isolation properties of virtual machines with the speed and efficiency of containers, doing so by implementing a minimalist architecture. The goal of this research work is to ascertain whether Firecracker meets the motivation for its creation and to measure the performance benefits of using Firecracker microVMs against more traditional virtualization platforms. Benchmarking tests are set up to measure application performance targeting various kernel subsystems. These experiments are then performed in multiple deployment environments: Linux machines sans virtualization, traditional VMS, Docker containers, and Firecracker microVMs.