Today, CITP has released the first update to the Interconnection Measurement Project website, consistent with the project’s goal to provide ongoing insights into the evolution of interconnection capacity and utilization, an important element of the Internet. This update includes three additional months of capacity and utilization data from participating ISP interconnection points. The visualizations now include the months of April, May, and June. The story remains much the same—the participating ISPs continue to add interconnection capacity to satisfy growing peak demand. The data shows this capacity growth in aggregate, across regions, and across participating ISPs. As of June 2016, aggregate peak utilization of interconnection capacity remains below 50 percent.
CITP works closely with CableLabs, Deepfield (the provider of the common measurement platform), and the participating ISPs to monitor the quality and integrity of the measurement data and to filter out measurement data that is not reflective of reality. From mid-April through the end of May, the measurement platform on a number of routers that account for a large portion of the interconnection capacity of the participating ISPs was being upgraded. Throughout that period, the measurement platform on these routers was providing unreliable data due to unanticipated consequences of that upgrade. In April, CITP received sufficient reliable data to generate an analysis on over 98 percent of the aggregate interconnection capacity of participating ISPs. Unfortunately, for May, we needed to filter the data for the entire month for almost 50 percent of the aggregate interconnection capacity; thus, the May analysis does not include those interconnects. We resolved these issues in time for the June analysis, which shows that over 97 percent of the interconnection capacity for the participating ISPs is included in the analysis. We’ve provided data for download where it is possible to examine these phenomena in more detail.
Because of the filtering, we omitted the regional views that New York and Los Angeles for May. We cannot show these regions because of the nature of our data-sharing agreement: because the number of participating ISPs included in the May analysis for these regions fell below three ISPs per region, the minimum level of aggregation for which we can release measurements under our data sharing agreement with the ISPs. In each of these cases, the data for the remaining ISPs in New York and Los Angeles became part of the “Others” category for May.
Additionally, in June, the number of participating ISPs in our analysis fell below three in the Phoenix region. The interconnection capacity that was unavailable in the Phoenix region for June was minimal, but nonetheless it reduced the number of participating ISPs below three in that region. Therefore, we also categorized the remaining Phoenix interconnect under “Other”.
CITP looks forward to building on this work and to providing a continued view of interconnection capacity and utilization. As always, we welcome additional participation from interconnection stakeholders to further broaden and deepen our research into this important aspect of Internet performance.