New ideas…
Proposed modern network architectures invalidate many of the assumptions of traditional networking. At the core of these assumptions is the structure of the modular, standardized protocol stack. Over the last ten years, the need for cross-layer design to meet new networking challenges has been a major research theme. Yet, there is also a recognition that the modular protocol stack has been a key enabler of the Internet; without a structure that allows the difficult problems in networks to be decomposed and decoupled, solving those problems might have been impossible.

The challenge, then, is to create new structures that preserve (some of) the modularity of the stack while enabling new types of inter-layer interactions and cross-layer controls. Thus far, though, no replacement for the traditional network stack has emerged.

Modifying the kernel…
Implementing experimental, comprehensive cross-layer approaches on commodity software and hardware remains challenging. Even the most accessible experimental networking platform — a PC running the Linux operating system — provides extremely limited transparency into network protocol performance. Gaining additional insight into protocol operation or implementing experimental network designs often requires the researcher to modify the operating system kernel, a task that is usually inaccessible for anyone except an experienced researcher. This dramatically limits the ability of undergraduate students, novice graduate students, and students enrolled in networking courses to “play” with the network stack. Such “play” is critical both for developing insight and understanding regarding protocol and network operation and for discovering new knowledge that is firmly rooted in the physical world (as opposed to knowledge based solely on analytical and simulation models).

In addition to system programming knowledge required to modify the kernel, changes to the kernel itself require the kernel to be recompiled, costing a minimum of 20 minutes, and the station to be rebooted each time.

The MANIAC Challenge…
Our most recent experience with experimental mobile ad-hoc network (MANET) research is the organization of the Mobile Ad hoc Networking Interoperability and Cooperation Challenge (MANIAC Challenge), an NSF-funded competition to better understand cooperation and interoperability in ad-hoc networks (the website can be found here). The following assumes that you are familiar with the MANIAC Challenge.

The MANIAC Challenge highlighted and revealed many of the shortcomings of current experimental platforms for wireless networking research. Some of the conclusions/by-products of the MANIAC Challenge were:

  • Having access to detailed data from all of the layers in the stack would have provided a much clearer picture of both the behavior of the entire network and the events and responses at individual nodes in the network.
  • Making changes to standardized networking operations is quite difficult. Non-standard routing behavior was required, but was difficult to implement. A “bubble gum and duct tape” solution was used.
  • The logistics of working with a medium-sized mobile wireless ad-hoc network were difficult; 15-20 nodes. Tasks like moving the laptops and even charging all of the nodes are difficult – weight, size, and electrical power quickly become issues. Distributing new code, collecting log files, and starting/stopping experiments takes coordination and effort.
  • If active participation in an experiment is not required, handheld devices can be activated and distributed to participants that can place them in a pocket or backpack and move at will throughout a predefined area while experiments are run on the network. If live network usage data is required, then participants can use the devices to access network services; even in this case, the handheld form factor facilitates active mobility, as opposed to a laptop computer, which is awkward to use in a truly mobile fashion because a participant can do little else while the experiments are taking place.

Experimental work…
Simulation and emulation can only go so far in evaluating the performance and feasibility ideas in networking research. In addition, there is a growing mistrust of simulation results, typically due to unrealistic simulation models and unrealistic test scenarios and parameters. Examples are the disk model of transmission range and node speeds that do not match any object on the planet. In the end, implementation and real-world testing should be the final stage of any idea’s lifecycle. Only with real-world testing can the performance truly be assessed. Unfortunately, many of the difficulties above deter researchers from truly validating and testing their ideas.

Cognitive networks…
More recently, we (DaSilva and MacKenzie) have participated in defining the nascent research topic of cognitive networks. The aim of this new work is, largely, to utilize ideas of self-organization (from MANETs) and cross-layer reconfiguration to optimize the end-to-end performance of networks with constituent nodes that are reconfigurable — often software defined radios (SDRs) utilizing dynamic spectrum access (DSA). While this research topic is young, we have already experienced emerging problems stemming from the lack of accessible experimental platforms. Since most DSA and cognitive radio work has started from the physical layer, available enabling research platforms typically lack a proper network stack altogether and do not provide reasonable support for even basic random-access MAC protocols. While we continue to anticipate the eventual emergence of more appropriate SDR platforms that allow for networking research, we believe that experimental work on reconfiguring and re-conceptualizing higher layers of the protocol stack can prepare us to use such platforms when they become available. The availability of a networking framework with a fully reconfigurable and transparent network stack will greatly enhance this work.

In closing…
Much of the motivation for this project arose from our own research work. We simply looked back (and forward) at our ideas and asked ourselves “how can we help make this happen or happen easier?”



Bucknell UniversityVirginia Tech WirelessVirginia Tech

This material is based upon work supported by the National Science Foundation under Grant No. 0916300. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

National Science Foundation
Copyright © 2019 FINS Framework