Summary

The paper defends the micro kernel idea that considered as inefficient and inflexible. Author describes that the micro kernel idea is not a problem, however its implementation is tricky. µ-kernal is not portable itself but it makes the whole system more portable. Author explains the ways to improve µ-kernel’s efficiency and performance.

Problem

This paper is a defense against the notion of µ-kernel inefficiency and inflexible performance. Paper is trying to achieve the implementation solution that can surely improve the performance of µ-kernel.

Contribution

  • Paper provides the brief understanding of operating system kernel and µ-kernel.
  • While suggesting the implementation of µ-kernel author assumes that security is required and page-based virtual memory is used. User can create subsystems that must be isolated from other subsystems and two subsystems can communicate with each other.
  • Author introduces three major components address space, threads and IPC.
  • Address space contains on master address space, which uses the physical memory.  I/O devices and threading is build on top of address spaces.
  • Each thread has unique ID and one thread can communicate with other using the inter-process communication.
  • Thread switching is processor architecture dependent.

Flaws

  • Paper does the performance evolution over the specific hardware, which should not be admissible.
  • Performance Comparison between different hardware should be defined.
  • System is not portable.

Extra Question

Adding additional layers such as improving security could affect the performance.

Reference

https://dl.acm.org/citation.cfm?id=224075