% -*- mode: latex; tex-main-file: "pospaper.tex" -*-

\section{Low-level Platform and APIs}
 
The low-level platform needs to provide OS support to the routing
protocols, including memory management and scheduling.  This can best
be provided by taking an existing stable open-source operating system
as the base development platform rather than developing a complete OS
from scratch, and we aim to support both FreeBSD and Linux as viable
XORP OSes.

Whilst the original Unix forwarding code can be used, our preferred
approach is to replace it with a modular forwarding path based on the
Click modular router architecture~\cite{click}.  Unlike at the higher
level, where we care strongly about robustness and extensibility, in
the forwarding path we care strongly about performance and
extensibility.  Robustness is still important, but solutions such as
memory protection that serve well for higher level functionality are
not acceptable in the forwarding path.  Robustness here comes through
simplicity, and through re-use of well tested and well understood
software.

Click unambiguously satisfies these requirements.  A Click forwarding path
is built at run-time from a large collection of very simple composible
software elements, configured using a declarative definition of the
desired plumbing, and organized around the flow of packets.

Click forwarding paths are extensible in two key ways:
\begin{itemize}
\item Existing defined elements can by interposed into a forwarding
path to add new functionality.
\item New Click elements can be created, loaded as kernel modules, and
then plumbed in to the forwarding path.
\end{itemize}
The ability to add new functionality into almost any place in the
forwarding path, even at runtime, is a very powerful extension
mechanism.  It's applicability for researchers should be obvious, but
this is also a very useful capability in a production router.  For
example, if a network administrator needs to keep some traffic
statistics on a very specific subset of the traffic, it is simple to
define a new classifier that matches that traffic and insert it into
the relevant place in the forwarding path where it can maintain the
relevant packet counts.

When we need to use hardware assistance to improve performance, Click
should aid the integration of such hardware.  It is a fairly natural
process whereby some parts of the forwarding patch and performed in
what to other Click elements look like just another element, but in
fact are a shim to special purpose hardware.  However our initial
implementation does not target advanced platforms due to our limited
resources.