On Sun, May 06, 2001 at 01:28:25PM -0400, Chuck Larrieu wrote:
> Without getting into the relative merits of router running open versus
> closed code, or the obvious cost issue, what would be the advantage of a
> Linux OS versus IOS?
Why not consider open versus closed source code? The public availability of
operating system source code is an enormous advantage that Linux systems
(and a variety of others like Mach and the FreeBSD/NetBSD/OpenBSD family)
have over their proprietary counterparts. Source availability enables rapid
development by allowing for higher-quality feedback from people who aren't
directly involved with development. Anyone, anywhere, can submit a patch to
the development team - as well as to the public - to correct a bug or add a
feature. This has proved invaluable in security circles; in some cases,
kernel-level fixes have been written and made available within two to three
hours of the discovery of a new security vulnerability.
Source availability allows for advanced troubleshooting in the event that
you trace your problem to an operating system bug. With proprietary
alternatives, your only recourse is to notify technical support and hope the
developers get around to fixing your bug before it's too late to matter.
The same reasoning applies to adding new features or customizations.
It is rapidly becoming clear that public availability of program source code
directly affects the quality of that code. Such availability effectively
distributes the 'development load' among many more people, with all the
attendant benefits that distributed processing implies.
Some other immediate benefits of using something like a Linux-based system
on router hardware include instant support for and compatibility with
existing OS file formats and filesystem types; a much greater ability to
'tune' your kernel image to your specific situation, providing decreased
image size and situationally-optimized performance; the potential for much
more advanced user interface features; and immediately available tools that
can be easily modified and cross-compiled to run on router hardware
(tcpdump, packet generators, netcat, intrusion detection utilities, ...).
> Doesn't the "OS" have to be an inherent part of the "IOS" in any case? I
> presume that Cisco boxes operate as do any Von Neuman based architectures,
> and that the IOS is really more an application that is loaded via the boot
> proms, where the "operating system" resides? Am I completely out of the
> water here?
I don't know if you're out of the water, but I was unable to make sense of
this paragraph. Perhaps you mean to ask about the difference between IOS
and other operating systems like Unix/Linux? In that case, there's really
no difference at all - IOS is an operating system like any other, although
more specialized than Unix. Unix does, however, separates the kernel
(low-level hardware support, core I/O, and processor and memory management)
from user-level applications such as the shell (CLI), shared libraries, and
daemon processes such as inetd (the Internet protocol super-server) and
cron. Because of its historically specialized nature, IOS melds 'kernel'
functionality with 'application' functionality.
Experience has shown that the modular design approach scales much better in
the long run.
> In raw terms of what is happening on a router, does a Linux based OS versus
> whatever the Cisco IOS is really matter? in terms of code size? In terms of
> router speed?
This is purely a 'one OS against another' issue. Is Windows 2000 'better'
than Linux if you have an Intel box? Despite the religious handwaving of
the advocacy-inclined, the fact is it depends on what you want to do.
In addition and again, having more than one alternative available has
historically proven vastly beneficial to hardware lifetime and acceptance.
> The IOS, as best I can guess, has it's roots in C.
It is written in C (and assembler), as is Linux.
> web link below, there aren't a lot of features in these Linux OS's either.
> I suppose over time that will be resolved, but at what cost in terms of OS
> image size?
As discussed earlier, image size is much less of a concern with Linux right
now than IOS. The ability to situationally optimize a given image allows
you to include exactly the features you need, contributing to efficiency in
space (image size and memory footprint) and time (performance).
Furthermore, the Linux architectural approach is modular. Most kernel
functions are now available as loadable modules which can be dynamically
loaded and unloaded during runtime.
> writing for a Cisco box, they have to ensure compatibility in every
> way shape and form with other Cisco boxes,
What kind of compatibility? Network protocol-wise? That's the reason why
standards and open specifications exist - they promote interoperability.
That's why, in a different OS implementation, you'll get OSPF and BGP, but
not EIGRP.
In the realm of the 'implementation dependent,' there are bound to be
interoperability issues. This transcends the 'Linux vs. IOS' question; it's
a general fact of "intercomputing." The Linux people are probably more
experienced at generating 'illicit compatibility' (interoperability in spite
of the unavailability of specifications, design documents or source code)
than anyone else.
--
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=3399&t=3362
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
