This paper distills some of the current thinking about Multicore OS scheduling into a set of design principles. This is good as most communications is bursty in nature, so how does an OS schedule the parallel applications on a multicore machine.
Section 3: Design Principles is hard to read, I would have thought that a list or table would have been easier to scan (of course I am scanning the articles). So if you are a little confused about the paragraph headings I would recommend that you cut and paste each paragraph and it will make more sense. I was able to make out the points and hope the authors will fix this in later pubs.