Sound travels 1000s of kilometres underwater; depending on its frequency, its variety of wavelengths enables probing of the ocean from millimeters to megameters. In this project, we resource the natural ambient sound as the probe with distributed sensing of optical fibres within legacy seafloor cables as vast arrays of passive acoustic receivers. The amplitude, phase and travel time of acoustic signals are strongly affected by the water temperature and flow velocity fields in their path. To obtain spatially resolved variability in these measurands, tomographic techniques can be used to combine integrals over several acoustic paths that connect a source and a receiver. Access to a higher number of acoustic paths improves estimation of ocean structure. Notable examples of oceanic phenomena already captured by tomographic techniques comprise convective chimneys in the Greenland Sea and basin-scale inversions of thermal structure. Despite these promising examples, use of active acoustic tomography is limited due to i) the economics of maintaining a powerful acoustic source (with noise-pollution consequences on marine life), and ii) the limitations on lateral and temporal resolutions associated with practical constraints on acoustic paths from active sources. Noise interferometry (NI) overcomes these limitations by replacing the use of active sources with diverse and broadband (10^-3 Hz - 10^-5 Hz) ambient marine noise, entails cross-correlating pressure fluctuations at different locations to retrieve an approximation to the acoustic Green's functions of various waves (i.e. the deterministic wave field due to a point source), which is then inverted to obtain ocean structure. This approach transforms any pair of discrete acoustic sensors (say, hydrophones) into virtual acoustic transceivers, which enables the quantification of both path-integrated sound speed (which is a function of temperature and pressure) and velocity. Flow velocity is retrieved from travel time nonreciprocity, i.e. the difference between travel times in opposite directions between two transceivers. Insensitivity of acoustic non-reciprocity to uncertainties in sound speed and transceiver positions enables accurate passive measurements of the oceanic current velocity, despite its absolute magnitude being less than the uncertainty in sound speed. When used with discrete sensors, NI requires maintaining sub-millisecond clock accuracy on underwater moorings for months-long periods and impractically large number of discrete sensors for useful spatio-temporal oceanographic measurements. This work overcomes these problems by replacing sparse point sensors (hydrophones/seismometers) with the data obtained using distributed sensing of optical fibres within offshore legacy seafloor cables. This enables spatially resolved O(10 m), dynamic measurements of relative deformation in optical fibre under the influence of ambient noise fields. Whilst these measurements are fundamentally different from acoustic pressure measured using conventional hydrophones, their sensitivity is comparable. In the NI context, the required time synchronization is greatly simplified as all signals come from the same fiber, with real-time data availability. Moreover, the large number of available sensor pairs and variety of pair-wise sensor separations yields a larger volume of input data for evaluating the noise cross-correlation function which results in the acoustic Green's function extraction, albeit with proportionately reduced noise averaging times, e.g., from hours-days to seconds-minutes. This project builds on the growing number of studies that have demonstrated the basics of the method by comparing inverse estimates from NI with directly measured time series of full ocean depth velocity and temperature. Our overarching aim is to determine the practical limits on spatio (vertical-horizontal) - temporal resolutions with measurand (temperature-velocity) precisions.

Internet eXchange Points (IXPs) have become a critical element of the Internet, as they provide the physical locations where networks interconnect and exchange traffic. IXPs carry huge traffic volumes, reduce interconnection costs, and hence make national Internet access affordable. Despite the growth of these infrastructures, the rapid evolution of the Internet poses new challenges. Reacting as soon as possible to the highly dynamic Internet environment has always been the first priority for Network Operators. Unfortunately, state-of-the-art techniques are extremely limited. Networks use the Border Gateway Protocol (BGP) to inform each other of which destinations are reachable. Accordingly, network operators (ab)use BGP Traffic Engineering (TE) to tweak traffic paths. TE is a network-management tool allowing a network to adapt events ranging from a change in customer location to mitigating dramatically large traffic outbursts of a malicious Distributed Denial of Service (DDoS) attack. However, BGP-TE lacks programmability and dynamism: once BGP preferences are set up, they cannot react in real-time to network events. With a high-fidelity measurement-focused approach, a network could implement more sophisticated traffic management techniques. For example, any network connected through an IXP must implement ingress traffic filtering to avoid receiving undesirable traffic (e.g., DDoS attacks or resulting from misconfigurations). However, correctly controlling ingress filters is complex. Thus, most IXP customers unrealistically expect the organisations originating the traffic to manage any problem. TE limitations result from the inability of current Internet monitoring techniques to cope with the wide range of granularities of network events. While control plane related events (those concerned with the selection of paths/routes, such as BGP updates) happen at a time scale of minutes, data plane events (packet processing) occur at time-scales of micro-seconds. While control plane monitoring is relatively easy, data plane observability is poor, relies on expensive equipment, and does not scale. EARL addresses this imbalance between the ability to observe control and data plane, and the consequent limits on the detection and reaction to network events. EARL is a novel integration of monitoring mechanisms and reactive network management. EARL enables a prompt reaction to network events with its Software Defined Networking (SDN) approach. Because of the IXP's central role on the Internet and the critical nature at the national level, we believe that they are the ideal place to explore EARL's ideas. We will demonstrate how measurement-assisted network management permits new Internet-wide services and, enables the provision of services hitherto considered impossible or too costly to deploy. Our goal for the EARL project is to pioneer SDN enabled measurement-based network management to enhance the Internet infrastructure. This will lead to relevant tools and data for the larger researcher and practitioner communities. To this aim, we will create a new research instrument, EARLnet: an operational, research-centered, Autonomous System (AS) directly connected to our partners, providing a new and unique real-world environment for the real-time monitoring of network status and SDN-oriented research. EARLnet will serve also as a test-bed to develop and evaluate novel reactive network management solutions. The EARL project has the potential to revolutionise current Internet network management through new fine-grained and reactive TE policies. EARL will not only create new mechanisms, but also translate the blind, legacy BGP-based, TE into measurement-assisted SDN techniques. Furthermore, through our partner institution, the Cambridge Cloud Cybercrime Centre (CCCC), EARLnet will provide valuable data to a large community of researchers and practitioners.