The Exokernel Operating System and Active Networks - Advanced Operating Systems and Kernel Applications - page 153

there is a separation of mechanism from policy.
Although the Exokernel has a policy of only
allowing code in user-space to handle manage-
ment of processes, perhaps on occasion the actual
mechanism of management has to allow for some
code to be executed within the kernel whenever
it is more efficient to do so.

Optimization will eliminate the design approach
seen in the current monolithic operating systems
and give birth to a new breed of operating sys-
tem that can keep pace with advancing computer
technology. The entire computer science industry
will experience a tremendous speed-up once the
extensible operating system design becomes fully
embraced.

optimizing usage

is extensibility the answer?

When we talk about optimizing a computer, we
are really talking about optimizing computer
hardware rather than computer code. Admittedly,
we do optimize computer code, but only as a way
to optimize computer hardware. Thus, optimiz-
ing computer hardware is always the real goal.
As such, who is the most qualified to optimize
computer hardware? Software engineers? One
might argue that hardware engineers - those that
have an intimate understanding of the hardware
components - are best suited for optimizing
computer hardware. If the extensible approach
becomes main-stream, we might see the line that
separates computer engineering from software
engineering becoming less distinct. Perhaps the
engineers of the future will play two roles - one
of hardware engineer and one of software engi-
neer. The engineers of the future will surely need
a strong understanding of computer code - since
that is how we communicate with hardware - but
they will also need an expert understanding of
computer hardware, since that is what is actually
being optimized. So, the extensible approach to
operating system design might lead to a paradigm
shift in how computers are optimized - but the
final result will be a society of fully optimized
computers.
The paradigm shift that will be caused by the
extensible operating system will help put the sci-
ence back into the Computer Science of the Infor-
mation Technology Industry. Computer developers
will be forced to make good Computer Science
decisions from the ground up. Optimization will
become the central focus of the computer industry.

Druschel and others have argued against the Ex-
okernel by saying “it is unclear to what extent the
performance gains are due to extensibility, rather
than merely resulting from optimizations that
could equally be applied to an operating system
that is not extensible” (Druschel, Pai, & Zwae-
nepoel, 1997). Through their research, Druschel
and others have shown that traditional monolithic
operating systems can be optimized just like the
Exokernel. They claim the key to the speed-up is
the optimization, not extensibility.
The Druschel research group tempers their
argument against the Exokernel by saying “the real
value in extensible kernels lies in their ability to
stimulate research by allowing rapid experimenta-
tion using general extensions” (Druschel, Pai, &
Zwaenepoel, 1997). They seem to be saying that
extensible operating systems provide a means to
quickly engineer prototypes of operating systems.
This fast prototyping has caused a speed-up in
the research, which had lead to a quicker way
to discover techniques for optimizing operating
systems. Although the Druschel group would say
extensibility is not the answer, they do support
the extensible approach because it is a tool that
can be used to speed-up research and help bring
a solution to market faster.
The argument provided by the Druschel group
is well founded. However, perhaps they are over-
stating their position. Although we will agree that
any code can be optimized - even monolithic op-
erating systems - we still hold strong to the point
that it is extensibility that really makes optimiza-