RSSAC047v2: RSSAC Advisory on Metrics for the DNS Root Servers and the Root Server System

The RSSAC recommends the ICANN Board commission an initial implementation of the measurement system described in this document to gather operational data and experience from actual monitoring of the RSS. The initial implementation should be designed such that it can transform into the official implementation as described in Recommendation 2 below. The insights learned from the implementation will inform future revisions of this document, if necessary.

The RSSAC recommends that the official implementation of the metric system must:

a. Meet the minimum requirements specified in Section 3 of this report regarding the number, location, connectivity, and other requirements for the vantage points.

b. Publish all software related to its operation under an open source license as defined by the Open Source Initiative.

c. Make the raw measurement data available to anyone in the interest of transparency. A third party should be able to use the raw data to verify the computation of these metrics.

d. In its monthly reports, only publish threshold pass or fail indicators for each RSI, not the actual measurements or metrics used to determine the threshold pass or fail values.

e. Publicly describe its methods for collecting measurements and aggregating metrics, including the topological location of each measurement vantage point. This description should be complete enough for RSOs and DNS researchers to create their own measurement collection systems similar to those used by the official implementation.

f. Share with an RSO the underlying measurements and metrics that resulted in failure any time an RSI fails to pass a threshold test. The shared measurements and metrics must include all measurements from around the time of failure and must include all measured values for all transports and address types.

The RSSAC, in collaboration with ICANN and the Internet community, should consider the following additional work:

• For a holistic view of RSS performance, it may be desirable or necessary to include measurements for all instances of each RSI. The only reasonable way to provide for such a view would be through self-reporting. In the future, it should be considered to have each RSO perform self-reporting of the defined metrics to eliminate uncertainty of components not under the RSO’s control, and it should probably be tied to an SLA including compensation for the RSO to implement.
• Create a reference data set.
• Explore the financial aspects of increased accountability and how it might relate to these metrics.
• Keeping with the provisions of RSSAC037 and RSSAC038 publish a document that advises any bodies created as part of the ongoing evolution of RSS governance on how they should interpret and act on data from the measurement systems.
• Investigate a better long-term plan for the location of the vantage points. Such a plan would distribute the vantage points by network topology instead of geographic location.
• Whereas the current work is based on a largely empirical model of the RSS, future versions of this document may want to take a more analytical and theoretical modeling approach.