Powering up collaboration

By on 10 Jul 2015

Category: Tech matters

Tags: , ,

Blog home

Atlas Greek Titan

Emile Aben, a system architect at RIPE NCC, recently published a blog post discussing how RIPE can improve its Atlas coverage by deploying more probes that target ‘eyeball’ networks – where users are – as opposed to core/infrastructure networks such as exchange points, deep inside an ISP’s own net, or at a Content Distribution Network or data center.

In this article he compares RIPE Atlas deployment against user population estimates provided by APNIC to see which eyeball networks are missing out on RIPE Atlas probes.

Collaborating on RIPE ATLAS

The project is (in part) one of the results of a five-month residency I’ve been having at RIPE NCC Labs, hosted by The Science Group Manager Robert Kisteleki. Emile and I have been involved in an ongoing project examining the convergence in measurement results from Atlas and the APNIC 1×1 ‘web blot’ system, and we have been discussing data sharing using the JSON format of data.

The APNIC Labs tests are held in an SQL database. By providing a simple URL name space for Emile to call through the web, we can “feed” him structured data of the APNIC view of relative market share, which is based on our random measurements per economy.

Given a good random distribution of tests, we get an insight into the relative market share of each economy from this sample. When we combine this data with ITU published statistics on Internet Subscribers and populations we can derive a first-approximation of the customers, subscribers, users in each economy by the origin-ASN.

Our data for populations is still indirect, and based on ITU collated statistics on subscribers along with other sources of world population, GDP and related data, but is based on a solid belief in the random data sample we see.

We have good reason to believe the advertising system continues to present us with unique random IP addresses, and we are not aware of any systematic bias ‘preferring’ specific sources (ISPs) of users who are given our adverts to view. We have presented the ranking data over several years to ISPs worldwide and received good feedback that the fundamentals of our relative market share projections are within the first approximation of their own view of the network.

I believe this is one of the only independent models of Internet relative market share available worldwide from the same technique, rather than the per-economy statistical reports, which are made periodically (infrequently).

At APNIC Labs we are actively sampling the userspace and drawing an inference from what we see.APNIC Labs has been providing data to the RIPE NCC for some time, as part of their IPv6 “stars” project. Discussions during my time in the Amsterdam office made it clear a JSON format data feed would permit far wider and simpler integration with their data analysis tools, and so we’ve done this ‘collaborative’ project to show (in part) what we can achieve together. We’re exploring other tools and data opportunities.

Reflecting and planning for the future

On reflection, it has been very pleasing to showcase RIPE NCC research in the APNIC region, and we expect more coverage of information from the RIPE Labs in the future. I have very high hopes we can continue to work from a distance.

The RIPE NCC R&D Group have been very kind and inclusive, and I have enjoyed my time working alongside them immensely, as well as other interactions with the Software Development, Policy, Root server/DNS and Communications groups. My stay was made much easier because of the assistance of the RIPE HR Team in particular Tineke Schmersel, Nicole Kelling and Daniella Coutinho from the Management Support group.  I also extend my thanks to all the RIPE NCC staff, and their Director Axel Pawlik for their hospitality.

As my residency draws to a close, I’ve started planning for my impending three-month residency in LACNIC (Montevideo). I’m hoping we can continue to grow collaborative outcomes between the RIRs and demonstrate other uses of our combined research capabilities. For instance, the ‘stars’ system can now easily be converted to using JSON format data from the same source.

Rate this article

The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.

Leave a Reply

Your email address will not be published. Required fields are marked *