Report on the IRTF meeting at IETF 84

The present note provides a short summary of the activities of the different IRTF’s Research Groups (RG) that held a meeting during the 84th IETF in Vancouver (July 2012).

IRTF Open Meeting (general meeting open to everybody)

Status of the IRTF:

Active Research Groups:
DTNRG – Delay-Tolerant Networking Research Group
ICCRG – Internet Congestion Control Research Group
CFRG – Crypto Forum Reseach Group

Research Groups with little activity:
ASRG – Anti-Spam Research Group
NCRG – Network Complexity Research Group
NMRG – Network Management Research Group
P2PRG – Peer-to-Peer Research Group
RRG – Routing Research Group
SAMRG – Scalable Adaptive Multicast Research Group

Research Groups recently closed:
MOBOPTS – Mobility Optimization Research Group
HIPRG – Host Identity Protocol Research Group

Research groups recently formed:
ICNRG – Information-Centric Networking Research Group

Proposed Research Groups:
SDNRG – Software Defined Networking Research Group (Proposed)
EERG – Energy Efficient Networks Research Group (Proposed)


ANRP – Applied Networking Research Prize:
Alberto Dainotti for his research into Internet communication disruptions due to filtering: Alberto Dainotti, Claudio Squarcella, Emile Aben, K.C. Claffy, Marco Chiesa, Michele Russo and Antonio Pescape. “Analysis of Country-wide Internet Outages Caused by Censorship.” In the Proceedings of ACM Internet Measurement Conference (IMC), November 2011, Berlin, Germany.

In the first months of 2011, Internet communications were disrupted in several North African countries in response to civilian protests and threats of civil war. In this paper we analyze episodes of these disruptions in two countries: Egypt and Libya. Our analysis relies on multiple sources of large-scale data already available to academic researchers: BGP interdomain routing control plane data; unsolicited data plane traffic to unassigned address space; active macroscopic traceroute measurements; RIR delegation files; and MaxMind’s geolocation database. We used the latter two data sets to determine which IP address ranges were allocated to entities within each country, and then mapped these IP addresses of inter- est to BGP-announced address ranges (prefixes) and origin ASes using publicly available BGP data repositories in the U.S. and Europe. We then analyzed observable activity related to these sets of prefixes and ASes throughout the censorship episodes. Using both control plane and data plane data sets in combination allowed us to narrow down which forms of Internet access disruption were implemented in a given region over time. Among other insights, we detected what we believe were Libya’s attempts to test firewall-based blocking before they executed more aggressive BGP-based disconnection. Our methodology could be used, and automated, to detect outages or similar macroscopically disruptive events in other geographic or topological regions.

DTNRG Report by Kevin Fall

The Research Group has been active for several years and has accomplished the design and implementation of several delay-tolerant protocols.

They held an interim meeting at Google making a point on the work accomplished and future research directions. Detailed content of the meeting can be found at: http://

The group will switch to working similarly to an IETF working group, meaning that milestones and deadlines will be set to produce documents/results. Current work concerns naming (i.e., developing a Node ID namespace looking as well multicast delivery models) and security (i.e., key management).


ICCRG – Internet Congestion Control Research Group

– Yuchung Cheng: “The GreatSnipe TCP congestion manager”

The GreatSnipe congestion manager is based on two main points: i) switch congestion control from the sender (e.g. youtube) to the receiver side; ii) since the client (receiver) has better knowledge of the local conditions congestion control should be performed on a per-client basis (and not per-flow).

– Grenville Armitage, Lawrence Stewart: “FreeBSD-based TCP CC research: ‘Delay-gradient’ TCP; Modular congestion control framework”

The talk presented an overview of the implementation of a modular congestion control framework in the FreeBSD protocol stack. Such framework allows to use several different congestion control algorithms at the same time on a per-socket basis. Using the aforementioned framework a delay-based congestion control algorithm is also presented. Such algorithm uses variations in the RTT to estimate the degree of congestion. The implementation is freely available at:

– Matt Mathis: “Update on TCP Laminar”

The presenter gave a short update on the Laminar TCP implementation for Linux kernel (v3.5). The implementation has been re-engineered and split in several subsystems, increasing modularity and support for future add-ons. Discussions on future research related to this topic will take place on the mailing list:

– Jim Gettys: “Areas where bufferbloat research is needed”

The talk tried to boost discussion on open research issues related to bufferbloat and how it is evolving. Open questions range from measurement aspects (can we use NTP to do passive measurements? Can we find “dark” buffers?), to algorithms (is CoDel a good solution? Can we find new AQM? Does packet aggregation affect AQM?), including the issue of whether or not new applications (e.g. netflix) have an impact.

ICNRG – Information-Centric Networking Research Group

The chairs of the research group will soon provide a list of the work items to focus on and milestones to structure the work. Most likely the group will initially work on a survey document as well as on a problem statement document.

The chairs propose to tackle the following technical areas: naming; routing; congestion control; caching; instrumentation & measurement; security. For all of them the group should try to look at near-term solutions, looking at existing IETF protocols, and long-term solutions most likely based on new architectures and protocols.

Concerning the caching issue it would be wise to have a look to “in-network storage” works (both research work and IETF activities).

– Stephen Farell: “Names with crypto in ICN” (

The talk focused on cryptography in ICN names for which old security models do not apply , because of the promiscuous caching context. The important point is the difference between data and objects, for which either the same naming should be used or a mapping between the different naming schemes should be provided, otherwise protocol development will be cumbersome.

– Ashok Narayanan: “Routing Protocol Formats” (

ICN objects are identified by global content names, which is a similar issue as CDNs willing to disseminate content availability information. The latter can be achieved using BGP (“Integrating
Routing with CDNs”, Field et al, IEEE Infocom NOMEN 2012), hence the talk explores the possibility to use BGP also in the ICN context by using a new NLRI (Network Layer Reachability Information).

– Matthias Waehllish: “Threats in Content-Centric Routing” ( (

The talk focused on a security threats analysis in the context of ICN. On the one hand, some threats are proposed (e.g., based on resource exhaustion or state decorrelation), on the other hand some experimental measurements are presented as well.
The RG will discuss on the mailing list whether or not to go for a detailed document describing security issues related to ICN architectures.

– Akbar Rahman: “ICN Scalability and Deployability”

The talk focused on the use of DHTs (Distributed Hash Tables) as routing approach. In particular a multi level DHT infrastructure is proposed, integrating name resolution and routing/forwarding.

– Suyong Eum: “Potential Based Routing for ICN”

The talk tackled the problem of how to locate content in ICN by using PBR (Potential Based Routing). Such lookup/routing infrastructure is based on a “potential” defined as a function of the distance (number of hops) and the expected quality of the content. Every node will have a per-content potential that will be used to route interest messages.

– Damien Saucez: “Congestion control and in-network caching” (

The talk explored in-path caching in ICN. It is known that popular content tends to be cached closer to the consumers while least popular content tend to be cached close to the source. Such behavior impacts content retrieval time and fairness, indeed congestion control is usually related to the RTT. The results of the analysis shows how the most popular content obtains more

– Pierluca Montessori: “a Network Resource Booking Algorithm” (

The talk presented REBOOK, which is a general resource reservation framework actually designed for a general context. Due to its generality the framework can also be applied to ICNs, which is shown in the talk. There is also a C/C++ based prototype that can run on both Linux and Microsoft platforms.

– Lichun Li: “Information-Centric Network in an ISP” (

How can ICN be deployed in ISPs? The talk discuss this question providing some examples and showing the benefits of in-network caching.

– Suyong Eum: “Announcements: ITU-T SG 13” (

ITU-T SG 13 Q21 is looking as well into data-aware networking. Document writing is ongoing and should be published end of 2013. There is no clear description of what will be the content. These documents will be available only to ITU-T members and this can results in limited access to information. The IRTF is not supposed to write standards, hence, there is no notion of liaison with other SDO (standard definition organizations) (with the trivial exception of the IETF). This situation translates in not having a clear vision of what the relation/collaboration with ITU-T will be.


SDNRG – Software Defined Networking Research Group (Proposed)

This is not yet an official Research Group. The Research group formation process consists in acting as a real RG for three meetings, then the IRTF chair will decide whether or not to grant the status of official RG. This is the first meeting.

– Bob Briscoe: “SDN: Solution-based or Problem-based Research?”

This was a general discussion on how to position SDNRG activities with respect to the general workflow of the IRTF. The IRTF is usually problem oriented, meaning that Research Groups focus on a specific problem to be solved (e.g., routing, DTNs, etc.). However, SDN is not a problem but rather a solution (framework). So what should the scope of the RG be? The suggestion is to have a solution-prompted approach, meaning to use SDN to re-evaluate/question traditional approaches (but without requiring that SDN must be the solution).

– Nick Feamster: “Applying SDN to Network Management Problems”

As the title clearly states the talk explored how to apply SDN for network management. The main assumption is that network management is all about event processing. In this context three approaches are presented: Lithium (event-based programming to express network policies); Bismark (programmable home routers); Procera (Functional Programming). Programmability is the key, hence future work will focus on languages.

– Edward Crabbe: “SDN at Google”

The talk overviews the SDN experience at Google and how they use it to optimize their WAN. In particular some benefits obtained by centralizing the traffic engineering. This is something actually deployed in Google’s WAN. Nevertheless, more has to be done to provide better abstraction between the controller and the physical switches, the controller and the applications, and define the interaction model between different controllers. Some of the aspects of these three points are explored.

– Myung-Ki Shin: “Formal Specification for Software-Defined Networks” (

The talk explored the issue of formal definition of SDN. After exploring different possibilities the use of three-tier architecture coupled with the ACSR (Algebra of Communicating Shard Resources) language is proposed. The three tiers are: i)
Forwarding entities; ii) Control and management entities; iii) Applications and Services using tier-1 and tier-2 entities. Any solution can match this three-tier architecture and be formally specified by using ACSR.

– Curt Beckman: “Southbound APIs and Protocols”

The talk is about the need of an abstraction layer between the SDN controller and the physical devices (the so called southbound API). Due to the fact that routing/forwarding devices are usually
heterogeneous, the central controller become overly complex if it has to deal with all the peculiaritities of each device, hence the need of an abstraction layer. Even OpenFlow failed so far in providing such simple layer.

– Diego R. Lopez: “SDNi: A Message Exchange Protocol for SDNs” (

The talk tackles the problem of SDN partitioning. In very big networks it is unthinkable to use one single SDN with a central controller. It is inevitable that big networks will be partitioned in smaller SDN domains. In a multi-domain SDN context an interface is needed between the different domains: SDNi.

Final discussion in the RG focused on the next steps. The main general goal of the group will be to “Evolve the Internet at large by exploring various SDN architectures”. This means to develop a work plan and a taxonomy, which will be done in the next months. There is already the idea to produce two documents:
– SDN Problem statement
– Survey on existing SDN models
The exact structure and content of the two documents is still under discussion.

EERG – Energy Efficient Networks Research Group (Proposed)

This is not yet an official Research Group. The Research group formation process consists in acting as a real RG for three meetings, then the IRTF chair will decide whether or not to grant the status of official RG.
This is the second meeting, the first being held during 82nd IETF in Quebec City.

Differently from the other RGs, this session did not include any presentation; every interested person can have a pitch of 5 minutes. The focus of the meeting was to structure the problem space and who is willing to work on it.


Samsung is interested in having a community where to exchange idea and real network data. They claim the effort will provide a good ROI (Return Of Investment).

Hitachi is interested in power consumption in data centers. More precisely, how to optimize VM allocations in order to reduce power consumption?

Huawei is looking in solution based on traffic aggregation in order to idle (and hence power-off or put in low consumption state) some of the devices of the network. The trade-off between energy efficiency and resiliency is something to be explored.

No slides are available but details about the subsequent open discussion can be found at:

[Report submitted by Luigi Iannone, ITC-IRTF liaison]