Last week, I had the pleasure of attending the 2016 Asian Internet Engineering Conference held in Bangkok, organized and supported by AIT, ThaiREN and WIDE.
This is the second time I’ve attended this conference, after attending a 2014 meeting held in Chiang Mai. This time, around 30 researchers met for a three-day ACM/IEEE-backed workshop to discuss their current work.
This is a very active and focused community, which includes Thai research groups, visiting scholars and researchers from the Asia Pacific, graduate and post-graduate students from the region, and invited guests from worldwide research and teaching communities. AINTEC is deliberately kept small as a workshop/meeting, and retains a friendly and approachable feel, with student poster sessions held alongside keynote talks.
It’s a great social mix and consistently very interesting. Amusingly for me, one lunch meeting turned out to have researchers from St Andrews, and Cambridge University in the UK, whom I have either met before in research colloquia or who work with people I know; a timely reminder of how remarkably small and well-connected the research community can be in our field. Here we all are, meeting in Thailand from all the corners of the Earth.
This year, several talks stood out in my mind, and I want to flag them and their speakers here.
Artificial Intelligence and information theory
Philippe Jaquet (Bell Labs Alcatel-Lucent) gave a very entertaining and erudite talk about AI and information theory. His presentation included an imagined conversation between Claude Shannon and Alan Turing (with interjections from Donald Knuth) about his research group’s work performing semantic analysis of tweets, in language neutral manners detecting tweet authorship and structure similarities in real-time.
Philippe made an intriguing observation about the neurone and synapse ‘bit’ densities of human and rat brains, compared to the equivalent bit density of the tweet stream, which now approaches it in scale, if not in quality.
Mobile behaviour
Kunwadee (Kay) Sripanidkulchai (IBM T.J. Watson Research Center) presented an analysis of mobile application network behaviour, which focused on e-health. She is using a framework for testing mobile applications (Android and iOS) and measuring network traffic, including TLS-protected traffic. This allows her to detect adware, tracking, and possible data exporting to the cloud.
Remarkably (and very worryingly) almost 50% of the top health apps worldwide, on either platform, appear to be using unprotected web traffic. This is very concerning as a public health privacy outcome. This work deserves a lot wider discussion for its social/privacy policy implications.
Kensuke Fukuda (National Institute of Informatics, Japan) presented a long-baseline survey of domestic mobile Internet usage in Japan.
This is an amazingly deep study of the traffic behaviours seen in Japan, conducted in the same way over many years. It can show the rise of public WiFi, the change in P2P traffic against the legal environment for content sharing, and the rise of Netflix and other sources of data. I believe the Japanese community will be reflecting on this data frequently, as they develop their networks.
Software Defined Networks
Kotaro Kataoka (Indian Institute of Technology, Hyderabad) gave two presentations.
During his first presentation, he discussed a model of how to manage secure communications using Software Defined Network (SDN) switches to isolate untrusted client devices until they are validated.
His second presentation was about the overall SDN architecture and how this can be analysed using classical computer science models of data to construct smaller (and quicker to calculate) forwarding information bases (FIB) to improve switch flow processing.
Kotaro is a regular at AINTEC meetings and has presented previously on student network designs using simple lightweight WiFi meshes. His work on SDNs explores the emerging space of agile networks, security and efficient packet forwarding.
Detecting international BGP routing detours
Anant Shah (Colorado State University) presented on detection of International BGP routing ‘detours’, which trombone through external (third) countries.
Anant provided a fascinating review of data being culled from BGP as a live feed, combined with external information from geo-tagging, and other databases combined with RIPE Atlas data.
Several talks included the use of RIPE Labs’ Atlas for measurement, either directly or indirectly, (confirmation of a working theory in another technique) which is a strong confirmation of how valuable the worldwide network research community finds the investment the RIPE NCC has made this platform.
Anant presented that many events are being caused by remarkably few players and can be related to few prefixes. It also demonstrates ‘flash’ events, which appear to show divergent routes for banks and strategically important networks occurring. Even though it’s unclear if this is caused by misconfiguration or something more malicious, it’s nice to be able to detect the events happening.
Anant’s work includes information being exported in JSON as a public benefit, to mark ASN economic activity. This work won the best paper award, which is hugely satisfying for all concerned.
Identity-Locator Separation
Ditchaphong Phoomikiattisak (GISTDA) a former student of St Andrews, now residing back in Thailand, presented his PhD work on “IP without IP addresses” (which from an Internet Number community perspective, is a potentially frightening topic!).
In reality, his work focuses on the Identity-Locator Separation problem (RFC6830), which is a really active field of research trying to expose the confusion caused by making Internet numbers serve both purposes: the unique identity of things on the network, and the address to find them. The roots of this problem lie deep in RFC2101, an early RFC by Jon Crowcroft in 1997, and more recently Lixia Xiang’s RFC4984, and the Brian Carpenter paper IP Addresses Considered Harmful [PDF 664 KB].
DItchaphong discussed his work on Identifier-Locator Network Protocol (ILNP), which remaps into the address a distinct locator to find things and a node identifier, which is not used for routing or topology, or even to identify an interface. ILNP is neutral to IPv4 and IPv6, but it’s a lot easier to make it work in IPv6 by treating the IPv6 address fields slightly differently, yet making no structural changes.
Localizing content
Arjuna Sathiaseelan (University of Cambridge), talked about “5G for All”, a high-frontier view of enabling access to the as-yet unconnected communities worldwide.
Arjuna is the research team leader of the “networking for development” group at Cambridge and has good evidence of the global digital divide being a problem in all regions, particularly Africa. The group is exploring new models of constructing networks, which will exploit gaps in used radio space to try and get to communities within reach of major population centres – whitespace radio in television spectrum spaces is a good example of this.
Arjun argued cogently for more localized data, including positioning CDNs closer to communities, particularly when they store content that is relevant and appropriate to the locality. His work is funded by significant public interest bodies and is going to be a huge commitment of investment to these communities if it comes to fruition.
Adisorn Lertsinsrubtavee, also from the University of Cambridge, gave a talk about name-based networking, IoT, and its application to models of publish-subscribe information flow and multicast routing.
The aim of such networking (which fits very nicely with Arjun’s work on localized and fast services on the edge of the network) is to be highly flexible in how “things” can be named, grouped and controlled, with efficient use of abstract command-response routing and efficient use of network resources.
A comment Adisorn made to me outside the presentation reiterated something a lot of people at this meeting were exploring, which is the idea that “names” and “addresses” are becoming much less important in the conversation of the future network. Instead, people want to talk about “locators” and “models of naming” much more than low-level packet identifiers that the number registry system maintains. Adisorn argued in his talk for a more fused sense of names, with hierarchical (and DNS-like) hash, and capability tagged behaviours, taking the best parts of each naming model under consideration.
Rehan Noordally presented a case study on Internet latency for Reunion Island.
As a resident of Reunion, Rehan shared with us how telecommunications on the island suffers physically from extreme isolation even though it has three fibre connections, almost all of which connect to European long-haul fibre nets. The bizarre thing is that three local islands have all got connectivity on a common fibre but aren’t yet peering, so traffic which is inherently local, is going via a European IX.
Rehan was able to show that the extra acquired delay was almost exclusively limited to local traffic. All long haul traffic actually winds up acquiring less delay overall – it still takes a long time, but the delay is primarily driven by real physical distance and light speed, rather than excess path length from “trombone” routes.
In summary
What I like overall about the AINTEC meeting is the straightforward open discussion that is encouraged and enjoyed by everyone.
Poster sessions with students from the Thai university/research sector discussed traffic analysis by neural nets, public transport mobile communication models, thread analysis in OSPF, IoT and website classification models.
Having said all the above the three things that stood out for me the most from this meeting were:
- The role of data collection and experimentation tools like Atlas is now a given. Researchers in many disparate communities are building out models of network behaviour that expect to be able to use and benefit from Atlas. The RIPE NCC should be very proud of what they have built, judged on this.
- Having come here to sing the praises of data sharing, APIs, and formats like JSON and CBOR, I find that lots of the research work is already being done using APIs (like the Atlas data API) including also publishing APIs of information built up in the research, and using formats like JSON as Anant has done in his diverse path research.
- It’s so pleasing that all the work done by Dr. Kanchana Kanchanasut and the AIT staff has paid off in developing such a prosperous research community. I saw people who are consistent attendees, people I remembered from my last visit, new faces who are moving through their early stage research careers, and older academics and researchers who enjoy the conversation and are coming back to refresh. For this reason, it was great to be able to celebrate Kanchana’s Jon Postel award, which she was awarded at IETF 97 last month.
I’ve recommended several talks to be repeated into the operations community in 2017, and I hope the work gets wider coverage and discussion.
I also hope I can come back here sometime in the future; it’s nice to renew relationships and explore the development of work like this.
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.
Thanks George . nice article with summary