CNTR: Lightweight OS Containers
CNTR: Lightweight OS Containers Jorg¨ Thalheim, Pramod Bhatotia Pedro Fonseca Baris Kasikci University of Edinburgh University of Washington University of Michigan Abstract fundamental to achieve high efficiency in virtualized datacenters and enables important use-cases, namely Container-based virtualization has become the de-facto just-in-time deployment of applications. Moreover, standard for deploying applications in data centers. containers significantly reduce operational costs through However, deployed containers frequently include a higher consolidation density and power minimization, wide-range of tools (e.g., debuggers) that are not required especially in multi-tenant environments. Because of all for applications in the common use-case, but they these advantages, it is no surprise that containers have seen are included for rare occasions such as in-production wide-spread adoption by industry, in many cases replacing debugging. As a consequence, containers are significantly altogether traditional virtualization solutions [17]. larger than necessary for the common case, thus increasing the build and deployment time. Despite being lightweight, deployed containers often include a wide-range of tools such as shells, editors, CNTR1 provides the performance benefits of lightweight coreutils, and package managers. These additional tools containers and the functionality of large containers by are usually not required for the application’s core function splitting the traditional container image into two parts: the — the common operational use-case — but they are “fat” image — containing the tools, and the “slim” image included for management, manual inspection, profiling, — containing the main application. At run-time, CNTR and debugging purposes [64]. In practice, this significantly allows the user to efficiently deploy the “slim” image and increases container size and, in turn, translates into then expand it with additional tools, when and if necessary, slower container deployment and inefficient datacenter by dynamically attaching the “fat” image.
[Show full text]