My major research areas of interest has evolved over the now many years of systems research, starting in early university times, leading me into mobile-centric research (e.g., at Nokia Research) over new network architecture (e.g., at BT and Cambridge University) to 5G (e.g., at InterDigital) and new Internet technologies (currently at Huawei Research).

Some of those areas include:

Routing on Service Addresses (ROSA)

Pursuing a different path from previous path-based forwarding research (see further down), Routing on Service Addresses (ROSA) pursues the vision of a service-centric delivery model at Layer 3.5, i.e., realized as a shim overlay atop IPv6. The proposed design utilizes IPv6 destination extension headers to route an initial service request directed to a service address across (overlay) service address routers (SARs). With service addresses underlying an anycast semantic, i.e., more than one service instance may realize the service exposed through the service address, it is the task of those SARs to select one of the possibly many instances in their routing and forwarding decisions. The selected instance then replies back to the client through the overlay, while continuing packets are now sent directly to the IPv6 address of the instance, thus providing affinity of packets within a longer service transaction. Through this, ROSA provides an alternative to the explicit resolution step in the DNS, replaced with an in-band discovery process, while utilizing IPv6 for all following requests within the transaction.

In the IETF draft on ROSA, you can find use cases, motivation, and benefits of such approach, together with an analysis of why existing network technologies do not fulfil the dynamicity requirements posed in those use cases.

ROSA has been implemented through an eBPF-based forwarding prototype, while utilizing a rawIP-based socket library for adapted clients and servers, utilizing the new ROSA capabilities. For more information, drop me an email!

Compute-Network Integration

The integration of computing and networking infrastructure is another key topic of my research. Exposing computational properties in order to improve on network-level traffic steering decision is key to many use cases, such as those in (5G) edge computing, vehicular computing or distributed AI, where sending to the ‘best’ endpoint in terms of shortest path (or lowest latency) does not necessary result in the best service performance, e.g., when the shortest distance endpoint is overloaded.

My contributions to work in the IETF are currently finding input into the newly formed Compute-Aware Traffic Steering (CATS) WG, where I have co-authored use case, requirements, gap analysis and initial framework drafts. Although this WG is new to the IETF, there has been preparatory research work, outlining the advantages of taking computational input into account for traffic steering decisions, such as the methods developed for our CArDS (compute-aware distributed scheduling) solution. When combined with the highly dynamic capabilities of ROSA (see above), traffic steering decisions can be as dynamic as per service transaction down to single packets sent in the network. Key here is the overlay nature of the steering decision maker (the SAR in the case of ROSA), allowing for separating the concerns of network and service provider in this new role of a service communication provider. Ongoing research is investigating the future opportunities that this new role may provide.

In-Network Computing

Another key direction of future networking technolgies is the execution of computational functions in the network, rather than at endpoints only (see this paper for a discussion on how this aspect may relate to the End-to-End principle as formulated by Salzer et al.). This capability, termed as in-network computing, has a home in the IRTF within the Computation-in-the-network (COIN) research group. I have made several contributions to this RG, such as contributing to the use case analysis as well as taxonomy and other aspects.

Distributed Consensus Systems (DCS)

Large-scale distributed consensus systems are often known as realizing, e.g., cryptocurrencies (e.g., in the form of blockchains) or distributed file storages. They utilize insights and mechanisms developed for what is often called Distributed Ledger Technologies (DLT), realizing a record keeping for a distributed computational function much akin to a distributed database. While much has been done to better understand the energy impacts of those DLTs, such as caused through proof-of-work techniques requiring significant computational, and thus energy, efforts, work to investigate the communication effort of the underlying protocols to disseminate information in those DCS/DLT platforms is relatively new. First published in an IIC whitepaper and later significantly expanded on in a separate paper, we provided insights on the impact of DLTs on transport provider networks by investigating the methods implemented at the many (hundreds of) thousands of peer nodes in those DCS platforms. We discovered and quantified the significant signaling costs for but also waste in communication that is taken place, most being driven by the needed consensus that those systems strive for, in turn leading to a recursive diffusion of information in the network. A key direction of ongoing research is to take those lessons learned to inform network-assisted methods, most notably based on ROSA methods (see above), that will not only help to drastically reduce communication costs in future DCS platforms but also improve on converging to a consented state much faster and thus reducing, possibly even removing the need for the many existing proof techniques that are often seen as the root cause for the negative environmental impact of DCS systems.

Service-based Architecture

During my time at InterDigital Europe in London, I developed initial solutions in the H2020 POINT and RIFE project that have then been further developed into a service-based architecture (SBA) platform with full ETSI NFV compliant orchestration of control and user plane services. In the H2020 FLAME project, we embed this platform into a future media internet delivery platform, together with advanced monitoring and control capabilities developed by IT Innovation. Trials and showcase efforts aim at developing use cases on future media delivery with user-facing delivery in smart city environments deployed in Bristol and Barcelona. We also deploy the platform in the UK-funded 5G Smart Tourism project with a final delivery planned for March 2019. The technologies developed for this platform are disseminated into standard organisations such as the IETF or 3GPP.

Service Routing based on Path-based Forwarding Methods

After joining InterDigital Europe in London in 2013, I started working on solutions that apply my work on information-centric networking in the context of flexible (service) routing. Within the EU-funded H2020 project POINT, we developed solutions for IP/HTTP-level flexible routing. These solutions provide significant performance benefits for operators deploying this solution in their networks, particularly SDN-enabled ones. This work is complemented by efforts within the H2020 project RIFE, where we develop solutions that lower costs of provisioning Internet access, striving for universal Internet access for everybody. In the two projects, I have been acting as Technical Manager (POINT) and architecture leader (RIFE), leading a small team at InterDigital towards developing and building the solutions designed within these projects.

Architectures and Technologies for the Future Internet

Since leaving BT Research in 2009, I focussed on Internet architecture related issues during my work at Cambridge University. The activities were a mix of technical but also community-related work.

Seamless and Context-aware Services

After joining Nokia Research in 2000, I started my work in the area of seamless and context-aware services. This topic comprises aspects to provide protocol and service functionality in a multi-dimensional heterogeneity, i.e., in the presence of heterogeneous access networks across operators and device boundaries. It further considers the adaptation of service and application functionality based on the currentcontext of the user. Context includes, but is not limited to, information about location, presence, current activity, affective state, or other information of the user.

Collaborative Computing

My initial research portfolio started during my PhD at RWTH Aachen, developing solutions for collaborative computing, specifically on high-speed transport protocols and conference control.