This packet format is called the “
Special-purpose virtual links in IP
networks are often called “tunnels.”
Our model provides a structured view
of tunnels, clarifying the roles of network members at the upper and lower
levels of tunnel endpoints, the state
that each network member requires,
and the fields that must be present in
packet headers. This uniformity across
levels can explain confusing designs
and make them analyzable. For example, even though a network can use
itself in a usage graph, a network link
must never use itself.
of the Compositional Model
Since 1993, the Internet has evolved
by means of new networks and new
compositions. The Internet today is a
vast collection of networks comprised
in a rich variety of ways by layering and
bridging, including being composed
with themselves. Networks are easy to
add locally (campus networks, cloud
computing) or at high levels of the approximate usage hierarchy (mobility,
distributed systems). They are slower
to disseminate when both global and
low in the hierarchy (IPv6).
This evolution, while necessary to
keep up with increased demand, new
technology, and many new requirements, has created tremendous complexity. First and foremost, our compositional model describes the current
complex Internet as precisely as the
classic Internet architecture described
the Internet of 1993. Because it is inherently modular, it also has the potential to organize, explain, and simplify
as well as to describe.
Based on our experience applying the
model to many kinds of networks and
aspects of networking, there are two primary reasons for adding a new network
to the global Internet architecture:
• The network provides a specialized
service or unusual cost/performance
trade-off through mechanisms that are
not compatible with the general-purpose classic Internet design.
• There is a need for two different
instances of a network structure with
two different purposes. As in LISP-MN,
member names might be either per-
manent identifiers or temporary loca-
tions. For another example, the topol-
ogy of a network might be dictated by
security partitions (VLANs) or by paths
through required middleboxes, as well
as by physical connectivity.
Layered networks hide information, which can make problem diagnosis very difficult.
19 On the other hand,
separation of concerns into different
networks is a way of taming complexity. This is especially obvious when networks are being added for the second
reason, and distinct topologies (for
example) are maintained by distinct
networks. Also, very often, it is more efficient to compose two networks than
to intertwine distinct structures in the
same network. This is illustrated well
by Qazi et al.,
16 which shows that the
conflation of a middlebox topology and
a physical topology would cause a combinatorial explosion of router state.
The most immediate potential
benefits of the new model are based
on its capacity to explain the complexity that is already present and
must be dealt with. The model can
be formalized through analytic tools
and reasoning technology, in support of robustness and verification of
trustworthy services. We also believe
the model should be used in gradu-ate-level teaching, to cover a wider variety of networks in a shorter period
of time, and to encourage recognition of patterns and principles.
Next, the model has the potential to
improve current design and development of software-defined networks.
Reusable patterns would both increase
the availability of different points in a
trade-off space, and make each easier
to deploy by means of reusable or generated software. Optimizations should become easier to apply, because the model can help us reason that they are safe.
Finally, a compositional model may
help us to find a simpler future Internet architecture that truly meets foreseeable requirements and might even
adapt to unforeseeable ones. Perhaps,
with study of compositional principles
and compositional reasoning, we can
discover optimal uses of composition,
in configurations that exploit its benefits and ameliorate its disadvantages.
This could be the basis of network architectures that offer both flexibility
and manageability. Pushing Internet
evolution in this direction would be a
truly worthy goal.
1. Andersen, D.G. Mayday: Distributed filtering for Internet
services. In Proceedings of the 4th USENIX Symposium
on Internet Technologies and Systems, 2013.
2. Andersen, D.G. Balakrishnan, H., Feamster, N.,
Koponen, T., Moon, D. and Shenker, S. Accountable
Internet Protocol (AIP). In Proceedings of ACM
3. Benson, T., Akella, A., Shaikh, A. and Sahu, S.
Clous-NaaS: A cloud networking platform for
enterprise applications. In Proceedings of the 2nd ACM
Symposium on Cloud Computing, 2011.
4. Blumenthal, M. S. and Clark, D. G. Rethinking the design
of the Internet: The end-to-end arguments vs. the
brave new world. ACM Trans. Internet Technology 1, 1
(Aug. 2001), 70-– 109.
5. Clark, D. D. The design philosophy of the DARPA
Internet protocols. In Proceedings of ACM SIGCOMM,
6. Clark, D.D., Wroclawski, J., Sollins, K.R. and Braden, R.
Tussle in cyberspace: Defining tomorrow’s Internet. IEEE/
ACM Trans. Networking 13, 3 (June 2005), 462–475.
7. Day, J. Patterns in Network Architecture: A Return to
Fundamentals. Prentice Hall, 2008.
8. Farinacci, D., Fuller, V., Meyer, D. and Lewis, D. The
Locator/ID Separation Protocol (LISP). IETF Request
for Comments 6830. (Jan. 2013).
9. Farinacci, D., Lewis, D., Meyer, D., and White, C. 2017.
LISP Mobile Node. IETF Network Working Group
Internet Draft draft-ietf-lisp-mn-04. (Oct. 2017).
10. Feamster, N., Rexford, J. and Zegura, E. The road to SDN:
An intellectual history of programmable networks. ACM
11. Fuller, V., Li, T., Yu, J. and Varadhan, K. Classless inter-
domain routing (CIDR): An address assignment and
aggregation strategy. IETF Network Working Group
Request for Comments 1519. (1993).
12. Handley, M. Why the Internet only just works. BT
Technology Journal 24, 3 (July 2006), 119–129.
13. ITU. Information Technology—Open Systems
Interconnection—Basic Reference Model: The basic
model. I TU-T Recommendation X.200. (1994).
14. Karsten, M., Keshav, S. and Prasad, S. An axiomatic
basis for communication. In Proceedings of HotNets-V,
15. Kim, C., Caesar, M. and Rexford, J. SEATTLE: A
scalable Ethernet architecture for large enterprises.
ACM Trans. Computer Systems 29, 1 (2011).
16. Qazi, Z.A., Tu, C. C., Chiang, L., Miao, R., Sekar, V. and Yu,
M. SIMPLE-fying middlebox policy enforcement using
SDN. In Proceedings of ACM SIGCOMM. 2013.
17. Roscoe, T. The end of Internet architecture. In
Proceedings of the 5th Workshop on Hot Topics in
18. Saltzer, J., Reed, D. and Clark, D.D. End-to-end
arguments in system design. ACM Trans. Computer
Systems 2, 4 (Nov. 1984), 277–288.
19. Spatscheck, O. Layers of success. IEEE Internet
Computing 17, 1 (2013), 3–6.
20. Venkataramani, A., Kurose, J. F., Raychaudhuri,
D., Nagaraja, K., Banerjee, S. and Mao, Z. M.
MobilityFirst: A mobility-centric and trustworthy
Internet architecture. ACM SIGCOMM Computer
Communication Review 44, 3 (July 2014), 74–80.
21. Wang, Y., Matta, I., Esposito, F. and Day, J. Introducing
ProtoRINA: A prototype for programming recursive-networking policies. ACM SIGCOMM Computer
Communications Review 44, 3 (July 2014).
22. Yu, M., Rexford, J., Sun, X., Rao, S. and Feamster, N.
A survey of virtual LAN usage in campus networks.
IEEE Communications 49, 7 (July 2011), 98–103.
23. Zave, P. and Rexford, J. The design space of network
mobility. In Recent Advances in Networking,
Olivier Bonaventure and Hamed Haddadi (Eds). ACM
24. Zave, P. and Rexford, J. Network Security; https://
25. Zhang, L., Afanasyev, A., Burke, J. and Jacobson, V.
Named Data Networking. ACM SIGCOMM Computer
Communication Review 44, 3 (July 2014), 66–73.
Pamela Zave ( Pamela@pamelazave.com) is a researcher
in the Department of Computer Science at Princeton
University, Princeton, NJ, USA.
Jennifer Rexford ( email@example.com) is the Gordon Y.S.
Wu Professor of Engineering in the Department of Computer
Science at Princeton University, Princeton, NJ, USA.
© 2019 ACM 0001-0782/19/3