This is definitely well worth the time to watch. The "dismantling of the Spin Lock Dispatcher" is particularly exciting and important. By the time Win7 is nearing end-of-support, it may not be at all uncommon to be using those 256 CPUs!
The discussion of MinWin hopefully will settle this issue once and for all (though I doubt it will). It is *NOT* server core (as Thurrott disputed recently after Russinovich's last attempt to explain it), and it is *NOT* a "new kernel" (as most everyone else seems to believe, including seemingly those that should be close enough to the source to know better--Mary J. Foley, etc.). It *IS* a "redistricting" (my word... for election season :P) of the existing kernel ("Cutler's NT," as Russinovich calls it... interesting to speculate on some of the dynamics of the core OS team here, since Cutler is now working on the cloud stuff) to isolate it as a discrete, atomic component of the OS. They've isolated a ~40 MB layer of the lowest level of the OS so far, and they will continue to break it out up the stack over time.
The upshot of all this? No... MinWin is not a "new, lightweight kernel" with "much better performance" (the NT kernel is not slow, anyhow). The project wasn't undertaken for the sake of performance--and how could it? It's the same kernel. The upshot is a more manageable codebase, and an easier time innovating and improving it in the future. (And as a side note to that point... the "Spin Lock Dispatcher" changes were made prior to MinWin's finalization.)
|