A late reply to your posting 7655:
"I was wondering if someone out there can help me compare NT to Unix.
I would think that UNIX as a program is more simple and efficient.
After all, it was written back in the days when computers had
more limitations. As this should be the case, how does the simplicity
in the UNIX code effect its performance? Can NT ever
expect to run as fast as UNIX, or would it take a rewrite to catch up
with UNIX? NT was written under less demanding circumstances and more
demanding circumstances depending on how you look at it. From the
standpoint of efficiency, I can imagine that alot of involved coding
went into NT because NT could afford to be more bulky than UNIX. On
the other hand, NT has been more demanding on the programmers because
they have Bill Gates (who I would think to be rather demanding)
waiting for them to finish so they can meet the expected release
date. Do you think these factors play a part in determining which
operating system is more efficient? What are the implications of
these factors on the larger networks in existence?"
NT and Unix cannot be compared without reference to the context
in which they are being measured. It is a trivial matter to find
cases where NT is better than Unix and vice versa. The question is
then, which contexts will become important in the future, and if
the operating system is important to that context, which OS is
preferable.
Simple arithmetic, or just referring to Gilder's law, states that
the utility of a network is proportional to the square of the number
of nodes. Very simply, networking will outstrip the utility of
standalone machines. Another analogy is specialisation in economics
where the nodes can be dedicated to the tasks in a network, being
more effective than each node replicating functionality.
On a network environment, the multiplicity of attached devices
raises a large numbers of operating systems and protocols. The
alternatives are a Tower of Babel of conflicting protocols,
a proprietory standard like NT which will only talk with others
who talks its protocols (PPTP etc), or open standards like TCP/IP.
Open standards have economic advantages over propriertory protocols.
It is relatively easy for 3rd parties to write to the protocol
which both removes the burden of development from a single source
and encourages specialisation of purpose. The danger is that
the barriers to market entry are significantly lowered and the
products become easily commoditised.
In regards to OS's in a networked environment, there are two
basic types, client side and server side. Neither NT nor Unix
can function effectively as real time OS's. They are too large
and too slow to handle asynchronous events as effectively as
nano-kernel solutions like QNX which run as low as 50K. On the
server side, NT will never be able to match Unix at the high
end, but this does not mean it cannot be significant in its
own right. Because MSFT controls the source code to NT, box
manufacturers cannot optimise the kernel effectively to the
hardware specifications to mimimise latency without MSFT's
blessing. But for MSFT to do so would allow NT to fragment
as Unix has in the past with applications being rewritten
for each hardware configuration to take advantage of the
hardware's idiosyncracies. On the other hand, this is exactly
Unix's strength in hardware specialisation.
Again, in a network environment, specialisation is the key,
this would favour Unix as the operating system of choice.
This will be more so as more devices are attached to the
internet as clients, requiring commensurately more powerful
servers. |
