Community-Oriented Network Measurement Infrastructure (CONMI) Workshop Report
ACM SIGCOMM Computer Communication Review (CCR), v.36, n.2, pp. 41-48, 2006.

[This report summarizes issues discussed at the first CONMI workshop held on 30 March 2005 in Boston, Massachusetts. Sponsored by the National Science Foundation's Office of Cyberinfrastructure (OCI-0532233), the workshop was intended to begin a discussion regarding the viability and utility of a community-oriented network measurement infrastructure. This report was published 20 December 2005 online at: http://www.caida.org/workshops/conmi/]

Up-to-date, relevant Internet research requires comprehensive network measurement, but conducting and supporting Internet measurement raises several daunting challenges for the research community and funding agencies. Researchers need current data to progress in many areas, including Internet topology structure, routing dynamics, security, and workload trends. Given the inherent diversity of the Internet, collection of data requires a large-scale, distributed network measurement infrastructure. However, several challenges must be solved to enable large-scale measurement: funding of collection infrastructure, preserving the privacy of ISPs and users, resolving legal and proprietary ownership concerns, and prohibiting experiments that might cause harm.

The Community-Oriented Network Measurement Infrastructure (CONMI) workshop brought together key members of the Internet measurement research community to discuss whether a community-oriented approach could address current and near-future challenges in large scale measurement. Our inspiration came from the astronomy and high-energy physics communities which have self-organized to build, operate, and allocate the use of large, unique, and expensive measurement platforms. The objective of this workshop was to explore whether this cooperative model would benefit the Internet measurement community.

Internet measurements must respect the privacy of both users and of network providers. We explored the privacy implications of network measurement, with particular attention to facilities that would accept experiments to be run on infrastructure deployed on actively used networks. Each experiment could be examined in advance through community mechanisms, e.g., a review panel, to ensure that the collection process was not harmful and that the results, if released, would not raise privacy concerns. We also explored whether a fundamentally new, community-oriented model for passive measurement could enable a dramatically more powerful set of measurement experiments. The workshop raised more questions than it provided answers regarding how to best use passive measurement infrastructure funding to support the needs of the network research community, and we expect this discussion to continue as progress is made in critical areas, especially the difficulties in funding evolving measurement equipment and balancing privacy and security concerns with collection of necessary, relevant trace data.

For active measurement infrastructure, development of a community-based measurement system seemed currently feasible. The primary considerations include how to coordinate measurement requests from a large community of researchers, how to ensure responsible use, and how to ensure integrity of the data if gathered by an unknown party. In both cases, some community-oriented program is likely to be necessary to accommodate as many needs of the community as possible as cost-effectively as possible. Given the limited funding available to invest in this kind of measurement infrastructure, an objective cost-benefit analysis of the payoffs is essential.