Network Arcitecture (R02) Papers

This course is mainly based around reading papers and learning how to find the positive and negative (in that order) in those papers.

Here's some jolly good advice on How to read a paper by Keshav from Waterloo, plus how to write a great paper and give a great talk about it by Simon Peyton-Jones, ex Microsoft.

We'll go through these topics at roughly one per week. There should also be some guest slots.

One thing I'd like readers to bear in mind is that one can take an evolutionary approach to network architecture change, or one can try to be revolutionary. In discussing a given paper, try to see which approach it is taking and whether this supports or undermines the viability of the proposed idea - this notion originated with Constantine Dovrolis and Jenifer Rexford in this nice counterpoint discussion. An important evolutionary refinement is Punctuated Equilibrium: which may be how technology (including networks) evolve really.

A very interesting complex systems/systems bio/eco/evolutionary view on how layered architectures evolve is this paper on Architecture, constraints, and behavior by John C. Doyle & Marie Cseteb.

Internet Architecture

This paper by Dave Clark of MIT is the starting point for all network architecture papers in form and content: The Design Philosophy of the DARPA Internet Protocols

Further background

  • Steve's (The Simple) Internet Protocol (SIP)
  • Paul's Internet Protocol
  • IPv6 is what is on the table but is it too little too late?
  • IETF history of IPNG for SIP, PIP and IPv6 disaster by committee

    Internet Alternative -- For presentation & discussion by you

    To contrast with unified view of network architecture, (which reuse the form, but have very different conclusions from the Clark Internet Architecture paper above), see these two papers

  • Haggle to contrast with "always on" view that IP takes, which one of you will cover.
  • Plutarch for a view of protocol plurality, which I'll talk about on the wednesday.

    Background for Essay 1, for nov 6

    See Essay 1

    Topology/Data Centers - Lecture & Talk

  • One of you will cover one of the Data Center architecture papers from these: Microsoft Azure, accelerators & pingmesh), versus Facebook and Google
  • I will cover Queue Jump (a paper)

    May also look at

  • Rack Scale Systems and/or Scale Out Numa Systems and/or Netmap and see also this super-accurate PCIe performance measurement paper.

    Background for Essay 2 for Dec 1

    See Essay 2

    ML and Networks

  • One of you can look at evolutionary programming for TCP and re-enforcement learning for cellular

    For the latter, see also Delay Tolerant Bulk Transfers, though manually scheduled more than ML...

  • I will cover Qjump and associated artefacts!

    2022 cornell sigcomm paper on Tbps to the host just to show where we are now with a "single"

    Mobile- Lecture & Talk

  • One of you will tackle this paper on fingerprinting Mobile Virtual Operators: Mobile Operator fingerprinting or the more recent Apps, Trackers, Privacy, and Regulators (your choice:)

  • I'll present some of the Wireless Capacity arguments,

    IoT Lecture & Talk

  • One of you can cover the IoT gap analysis or IoT priv attacks
  • I'll cover Connectivity, Mobility and Identifiers : Jon- Background:-

    See also IEN 1 which kind of covered this in 1977! And using ICN for mobility (a bit using like multicast), see Map-Me See also notes on IPv6 deployment challenges and Intentional Naming

    Transport - End-to-End -- Lecture&Talk

  • One of you can cover Transport Services shims: QUIC and this new IMC paper and SPDY
  • Then I will talk about CCN/NDN and related new architecturework viz NDN and RMI

    See also Audio Conference Tool Over Named Data Networking

    Cascades, Enclaves and AI/ML

  • And One of you can look at buzz traq or cascading route failures

  • One of you can cover SCONE

    Wrapup, and Background for Essay 3 for Jan 16, 2018

    See Essay 3

  • What is Network Architecture?
  • I will talk about challenges challenges challenges in multicast, CCNx, IPv6, multipath, and re-decentralisation and wrap-up!