Scaling a Microkernel-Based Operating System to Cloud Environments

Digital services are increasingly hosted by dedicated companies (also
known as cloud providers), instead of running locally with the end user.
Currently, the operating systems used for driving server-grade hardware
in cloud data centers follow a monolithic approach. While offering good
performance, monolithic operating systems expose a large attack surface
as they implement many OS features in the kernel. Microkernel-based
operating systems instead offer a much more secure architecture by
design. However, such systems are often optimized to run on small
computing platforms. Hence, when used to drive server hardware,
microkernel systems face deficits, particularly from a performance
perspective.

Starting with a presentation of existing work on hardening operating
systems used in cloud settings, this talk discusses the challenges and
opportunities of deploying microkernels in cloud environments, using the
design of container engines as an example. The talk furthermore gives
insight into approaches for deploying microkernel-based systems on
server hardware. It concludes with a short description of our efforts on
increasing the scalability of L4Re, a microkernel-based operating system
developed at TU Dresden.