Exokernels Anyone? 12
shankark asks: "I was doing a paper review on the exokernel, and found some pretty good design ideas in it comparing the work with the monolitic kernel and microkernel architectures. There was also a Slashdot story way back in 1998, but the thread sort of died out. I wanted to know if there is any work being done in trying to incorporate the exokernel design ideas into the mainstream linux kernel. What would be the pros and cons of doing such a thing?"
the MIT project (Score:4, Insightful)
But, is this really all true? I have some doubts, but it looks really interesting.
A microkernel. Wonderful --- not. (Score:3, Insightful)
Somehow, I really don't think that'll work out for a large operating system. (Mind you, for a box with simple functionality, this may actually be good in terms of performance.)
It seems like the only real argument they are pushing is speed. (But, although they don't really admit it, you put some major kinks in compatibility.) The only real comparable example is something called "Socket". Their version is twice as fast as a version running under BSD.
But then again, all the OS code that would make it 'play nice' with other things on the operating system probably aren't there, which would explain the gain. (I'll go ahead and throw out the notion of academic dishonest, which was there in microkernels, using different compiling options, etc. Ask Linux all about what he thinks of it.)
Anyhow, what little I've seen emphasis speed, but doesn't show the downside. That would appear to be the cost of converting *complex* applications, and the ability to coexist with other things.
I see no real benefit of going with this method. For example, if I was a company, given the choice of having my internal app ported to an Exokernel, or throwing more hardware at it, I'd throw more hardware at it. No brainer. No big boon for developers.