Towards a (Practical) Convergence of Multiple Network Architectures
by Klaus Wehrle

Abstract:

In recent years, many ideas for new network architectures and overlay
networks arose (e.g. RON, i3, HIP) in order to add new functionality to
the Internet or to cope with its limitations. Nonetheless, many of these

In recent years, many ideas for new network architectures and overlay
networks arose (e.g. RON, i3, HIP) in order to add new functionality to
the Internet or to cope with its limitations. Nonetheless, many of these
concepts are just patchwork to temporally cope with some limitations of
the Internet architecture and to provide needed functionality. These
patchwork solutions cannot offer the quality and efficiency of
in-network deployed services.

Currently, the research community and some network operators see a clear
need to address these limitations and to provide efficient in-network
services. Nonetheless, network providers, operators and users will adapt
to these changes at different speeds and to various extends.  As result,
from our point of view, there will be no single future Internet
architecture or single successor to the Internet  and it will not be
enabled over night. The world of communications will be a plethora of
different networks, even more than today. Our task will be to bridge
these networks and to make them as interoperable and as unified (from
the users perspective) as possible. Therefore, we identify the
following challenges:

- How can different (and not yet existing) network architectures be
bridged and made interoperable?

- How can users access different types of networks, preferably via one API?

- How can we guarantee that a network deployment can be easily adapted
even to the n+1th network architecture?

For overlays and new network architectures to gain real user acceptance,
there must be an easy way for migration. Users should be able to
leverage new network functionality without any modifications to their
applications and operating systems.

With OCALA (Overlay Convergence Architecture for Legacy Applications 
NSDI06), we already presented an architecture that achieved these goals
for overlay networks in the Internet. We think that the concepts of
OCALA are also a good starting point for the convergence of multiple
heterogeneous network architectures. It also allows legacy applications
to leverage the functionality of new architectures and overlays without
any source code modification, recompilation or reconfiguration. OCALA
allows a user to simultaneously access multiple overlays for different
heterogeneous network architectures. It also allows legacy applications
to leverage the functionality of new architectures and overlays without
any source code modification, recompilation or reconfiguration. OCALA
allows a user to simultaneously access multiple overlays for different
purposes, stitch together the varied functionality offered by different
overlays, as well as communicate with hosts residing in overlays the
user is not directly connected to. OCALA also factors out the common
requirements for supporting legacy applications  researchers can
concentrate on designing new network architectures, rather than getting
legacy applications to work.