Searcharxiv⌕ Search

arXiv subjects

Dimitra Simeonidou

Publications and source records attributed to Dimitra Simeonidou.

At least 37 records · Page 2Linked to original sources
arXiv2024

Profiling AI Models: Towards Efficient Computation Offloading in Heterogeneous Edge AI Systems

The rapid growth of end-user AI applications, such as computer vision and generative AI, has led to immense data and processing demands often exceeding user devices' capabilities. Edge AI addresses this by offloading computation to the network edge, crucial for future services in 6G networks. However, it faces challenges such as limited resources during simultaneous offloads and the unrealistic assumption of homogeneous system architecture. To address these, we propose a research roadmap focused on profiling AI models, capturing data about model types, hyperparameters, and underlying hardware to predict resource utilisation and task completion time. Initial experiments with over 3,000 runs show promise in optimising resource allocation and enhancing Edge AI performance.

cs.LG↗
arXiv2024

Federated Fairness Analytics: Quantifying Fairness in Federated Learning

Federated Learning (FL) is a privacy-enhancing technology for distributed ML. By training models locally and aggregating updates - a federation learns together, while bypassing centralised data collection. FL is increasingly popular in healthcare, finance and personal computing. However, it inherits fairness challenges from classical ML and introduces new ones, resulting from differences in data quality, client participation, communication constraints, aggregation methods and underlying hardware. Fairness remains an unresolved issue in FL and the community has identified an absence of succinct definitions and metrics to quantify fairness; to address this, we propose Federated Fairness Analytics - a methodology for measuring fairness. Our definition of fairness comprises four notions with novel, corresponding metrics. They are symptomatically defined and leverage techniques originating from XAI, cooperative game-theory and networking engineering. We tested a range of experimental settings, varying the FL approach, ML task and data settings. The results show that statistical heterogeneity and client participation affect fairness and fairness conscious approaches such as Ditto and q-FedAvg marginally improve fairness-performance trade-offs. Using our techniques, FL practitioners can uncover previously unobtainable insights into their system's fairness, at differing levels of granularity in order to address fairness challenges in FL. We have open-sourced our work at: https://github.com/oscardilley/federated-fairness.

cs.LG↗
arXiv2024

First Demonstration of 25λ x 10 Gb/s C+L Band Classical / DV-QKD Co-Existence Over Single Bidirectional Fiber Link

As quantum key distribution has reached the maturity level for practical deployment, questions about the co-integration with existing classical communication systems are of utmost importance. To this end we demonstrate how the co-propagation of classical and quantum signals can benefit from the development of novel hollow-core fibers. We demonstrate a secure key rate of 330 bit/s for a quantum channel at 1538 nm in the presence of 25 x 10 Gb/s classical channels, transmitted at an aggregated launch power of 12 dBm, spanning over the C+L-band in the same hollow-core fiber link. Furthermore, we show the co-integration of the classical key-distillation channel onto this fiber link, turning it into a bidirectional fiber link and thereby mitigating the need for multiple fibers. We believe this to be an important step towards the deployment and integration of hollow-core fibers together with DV-QKD for the inherently secure telecom network of the future.

quant-ph↗
arXiv2024

AI Model Placement for 6G Networks under Epistemic Uncertainty Estimation

The adoption of Artificial Intelligence (AI) based Virtual Network Functions (VNFs) has witnessed significant growth, posing a critical challenge in orchestrating AI models within next-generation 6G networks. Finding optimal AI model placement is significantly more challenging than placing traditional software-based VNFs, due to the introduction of numerous uncertain factors by AI models, such as varying computing resource consumption, dynamic storage requirements, and changing model performance. To address the AI model placement problem under uncertainties, this paper presents a novel approach employing a sequence-to-sequence (S2S) neural network which considers uncertainty estimations. The S2S model, characterized by its encoding-decoding architecture, is designed to take the service chain with a number of AI models as input and produce the corresponding placement of each AI model. To address the introduced uncertainties, our methodology incorporates the orthonormal certificate module for uncertainty estimation and utilizes fuzzy logic for uncertainty representation, thereby enhancing the capabilities of the S2S model. Experiments demonstrate that the proposed method achieves competitive results across diverse AI model profiles, network environments, and service chain requests.

cs.NI↗
arXiv2024

NetMind: Adaptive RAN Baseband Function Placement by GCN Encoding and Maze-solving DRL

The disaggregated and hierarchical architecture of advanced RAN presents significant challenges in efficiently placing baseband functions and user plane functions in conjunction with Multi-Access Edge Computing (MEC) to accommodate diverse 5G services. Therefore, this paper proposes a novel approach NetMind, which leverages Deep Reinforcement Learning (DRL) to determine the function placement strategies in RANs with diverse topologies, aiming at minimizing power consumption. NetMind formulates the function placement problem as a maze-solving task, enabling a Markov Decision Process with standardized action space scales across different networks. Additionally, a Graph Convolutional Network (GCN) based encoding mechanism is introduced, allowing features from different networks to be aggregated into a single RL agent. That facilitates the RL agent's generalization capability and minimizes the negative impact of retraining on power consumption. In an example with three sub-networks, NetMind achieves comparable performance to traditional methods that require a dedicated DRL agent for each network, resulting in a 70% reduction in training costs. Furthermore, it demonstrates a substantial 32.76% improvement in power savings and a 41.67% increase in service stability compared to benchmarks from the existing literature.

cs.NI↗
arXiv2024

Federated Analytics for 6G Networks: Applications, Challenges, and Opportunities

Extensive research is underway to meet the hyper-connectivity demands of 6G networks, driven by applications like XR/VR and holographic communications, which generate substantial data requiring network-based processing, transmission, and analysis. However, adhering to diverse data privacy and security policies in the anticipated multi-domain, multi-tenancy scenarios of 6G presents a significant challenge. Federated Analytics (FA) emerges as a promising distributed computing paradigm, enabling collaborative data value generation while preserving privacy and reducing communication overhead. FA applies big data principles to manage and secure distributed heterogeneous networks, improving performance, reliability, visibility, and security without compromising data confidentiality. This paper provides a comprehensive overview of potential FA applications, domains, and types in 6G networks, elucidating analysis methods, techniques, and queries. It explores complementary approaches to enhance privacy and security in 6G networks alongside FA and discusses the challenges and prerequisites for successful FA implementation. Additionally, distinctions between FA and Federated Learning are drawn, highlighting their synergistic potential through a network orchestration scenario.

eess.SY↗
arXiv2023

iOn-Profiler: intelligent Online multi-objective VNF Profiling with Reinforcement Learning

Leveraging the potential of Virtualised Network Functions (VNFs) requires a clear understanding of the link between resource consumption and performance. The current state of the art tries to do that by utilising Machine Learning (ML) and specifically Supervised Learning (SL) models for given network environments and VNF types assuming single-objective optimisation targets. Taking a different approach poses a novel VNF profiler optimising multi-resource type allocation and performance objectives using adapted Reinforcement Learning (RL). Our approach can meet Key Performance Indicator (KPI) targets while minimising multi-resource type consumption and optimising the VNF output rate compared to existing single-objective solutions. Our experimental evaluation with three real-world VNF types over a total of 39 study scenarios (13 per VNF), for three resource types (virtual CPU, memory, and network link capacity), verifies the accuracy of resource allocation predictions and corresponding successful profiling decisions via a benchmark comparison between our RL model and SL models. We also conduct a complementary exhaustive search-space study revealing that different resources impact performance in varying ways per VNF type, implying the necessity of multi-objective optimisation, individualised examination per VNF type, and adaptable online profile learning, such as with the autonomous online learning approach of iOn-Profiler.

cs.NI↗
arXiv2023

DRL-based Energy-Efficient Baseband Function Deployments for Service-Oriented Open RAN

Open Radio Access Network (Open RAN) has gained tremendous attention from industry and academia with decentralized baseband functions across multiple processing units located at different places. However, the ever-expanding scope of RANs, along with fluctuations in resource utilization across different locations and timeframes, necessitates the implementation of robust function management policies to minimize network energy consumption. Most recently developed strategies neglected the activation time and the required energy for the server activation process, while this process could offset the potential energy savings gained from server hibernation. Furthermore, user plane functions, which can be deployed on edge computing servers to provide low-latency services, have not been sufficiently considered. In this paper, a multi-agent deep reinforcement learning (DRL) based function deployment algorithm, coupled with a heuristic method, has been developed to minimize energy consumption while fulfilling multiple requests and adhering to latency and resource constraints. In an 8-MEC network, the DRL-based solution approaches the performance of the benchmark while offering up to 51% energy savings compared to existing approaches. In a larger network of 14-MEC, it maintains a 38% energy-saving advantage and ensures real-time response capabilities. Furthermore, this paper prototypes an Open RAN testbed to verify the feasibility of the proposed solution.

cs.NI↗
arXiv2023

Entanglement distribution quantum networking within deployed telecommunications fibre-optic infrastructure

Quantum networks have been shown to connect users with full-mesh topologies without trusted nodes. We present advancements on our scalable polarisation entanglement-based quantum network testbed, which has the ability to perform protocols beyond simple quantum key distribution. Our approach utilises wavelength multiplexing, which is ideal for quantum networks across local metropolitan areas due to the ease of connecting additional users to the network without increasing the resource requirements per user. We show a 10 user fully connected quantum network with metropolitan scale deployed fibre links, demonstrating polarisation stability and the ability to generate secret keys over a period of 10.8 days with a network wide average-effective secret key rate of 3.38 bps.

quant-ph↗
arXiv2022

Dynamic DV-QKD Networking in Fully-Meshed Software-Defined Optical Networks

We demonstrate for the first time a four-node trusted-node-free metro network configuration with dynamic discrete-variable quantum key distribution DV-QKD networking capabilities across four optical network nodes. The network allows the dynamic deployment of any QKD link between two nodes of the network, while a QKD-aware centralised software-defined networking (SDN) controller is utilised to provide dynamicity in switching and rerouting. The feasibility of coexisting a quantum channel with carrier-grade classical optical channels where both the quantum and classical channels are in the C-band over field-deployed metropolitan networks and laboratory-based fibres (<10km) is experimentally explored in terms of achievable quantum bit error rate, secret key rate as well as classical signal bit error rate. Moreover, coexistence analysis over multi-hops configuration using different switching scenarios is also presented. The secret key rate dropped 43% when coexisting one classical channel with 150 GHz spacing from the quantum channel for multiple links. This is due to the noise leakage from the Raman scattering into the 100 GHz bandwidth of the internal filter of the Bob DV-QKD unit. When coexisting four classical channels with 150 GHz spacing between the quantum and the nearest classical channel, the quantum channel deteriorates faster due to the combination of Raman noise, other nonlinearities and high aggregated launch power causing the QBER value to exceed the threshold of 6% leading the SKR to reach a value of zero bps at a launch power of 7 dB per channel. Furthermore, the coexistence of a quantum channel and six classical channels through a field-deployed fibre test network is examined.

quant-ph↗
arXiv2022

DV-QKD Coexistence With 1.6 Tbps Classical Channels Over Hollow Core Fibre

The feasibility of coexisting a quantum channel with carrier-grade classical optical channels over Hollow Core Nested Antiresonant Nodeless Fibre (HC-NANF) is experimentally explored for the first time in terms of achievable quantum bit error rate (QBER), secret key rate (SKR) as well as classical signal bit error rates (BER). A coexistence transmission of 1.6 Tbps is achieved for the classical channels simultaneously with a quantum channel over a 2 km-long HC-NANF with a total coexistence power of 0 dBm. To find the best and worst wavelength position for the classical channels, we simulated different classical channels bands with different spacing between the quantum and classical channels considering the crosstalk generated from both Raman scattering and four-wave-mixing (FWM) on the quantum channel. Following our simulation, we numerically estimate the best (Raman spectrum dip) and worst locations (Raman spectrum peak) of the classical channel with respect to its impact on the performance on the quantum channel in terms of SKR and QBER. We further implemented a testbed to experimentally test both single mode fibre (SMF) and HC-NANF in the best and worst-case scenarios. In the best-case scenario, the spacing between quantum and classical is 200 GHz (1.6 nm) with 50 GHz (0.4 nm) spacing between each classical channel. The SKR was preserved without any noticeable changes when coexisting the quantum channel with eight classical channels at 0 dBm total coexistence power in HC-NANF compared to a significant drop of 73% when using SMF at -24 dBm total coexistence power which is 250 times lower than the power used in HC-NANF. In the worst-case scenario using the same powers, and with 1 THz (8 nm) spacing between quantum and classical channels, the SKR dropped 10% using the HC-NANF, whereas in the SMF the SKR plummeted to zero.

quant-ph↗
arXiv2019

End-to-end Quantum Secured Inter-Domain 5G Service Orchestration Over Dynamically Switched Flex-Grid Optical Networks Enabled by a q-ROADM

Dynamic and flexible optical networking enabled by NFV and SDN are the key technology enablers for supporting the dynamicity and bandwidth requirements of emerging 5G network services. To achieve the objective of 5G, Network Services (NSes) must be often deployed transparently over multiple administrative and technological domains. Such case often presents security risks since a typical NS may comprise a chain of network functions, each executed in different remote locations, and tampering within the network infrastructure may compromise their communication. To avoid such threats, QKD has been identified and proposed as a future-proof method immune to any algorithmic cryptanalysis based on quantum-physics mechanisms. The maturity of QKD has enabled the R&D of quantum networks coexisting with optical networks using telecom equipment. This makes the QKD a suitable candidate for the security of distributed and virtualised network services. In this paper, for the first time, we propose a dynamic quantum-secured optical network for supporting network services that are dynamically created by chaining VNF over multiple network domains. This work includes a new quantum-ROADM, extensions to SDN-enabled optical control plane, and extensions to NFV orchestration to achieve quantum-aware, on-demand chaining of VNFs. The experimental results verify the capability of routing quantum and classical data channels both individually and dynamically over shared fibre links. Moreover, quantum secured chaining of VNFs in 5G networks is experimentally demonstrated via interconnecting four autonomous 5G islands simultaneously through the q-ROADM with eight optical channels using the 5GUK Exchange orchestration platform. The experimental scenarios and results confirm the benefit of the proposed data plane architecture and control/management plane framework.

cs.NI↗
arXiv2019

Atomic-SDN: Is Synchronous Flooding the Solution to Software-Defined Networking in IoT?

The adoption of Software Defined Networking (SDN) within traditional networks has provided operators the ability to manage diverse resources and easily reconfigure networks as requirements change. Recent research has extended this concept to IEEE 802.15.4 low-power wireless networks, which form a key component of the Internet of Things (IoT). However, the multiple traffic patterns necessary for SDN control makes it difficult to apply this approach to these highly challenging environments. This paper presents Atomic-SDN, a highly reliable and low-latency solution for SDN in low-power wireless. Atomic-SDN introduces a novel Synchronous Flooding (SF) architecture capable of dynamically configuring SF protocols to satisfy complex SDN control requirements, and draws from the authors' previous experiences in the IEEE EWSN Dependability Competition: where SF solutions have consistently outperformed other entries. Using this approach, Atomic-SDN presents considerable performance gains over other SDN implementations for low-power IoT networks. We evaluate Atomic-SDN through simulation and experimentation, and show how utilizing SF techniques provides latency and reliability guarantees to SDN control operations as the local mesh scales. We compare Atomic-SDN against other SDN implementations based on the IEEE 802.15.4 network stack, and establish that Atomic-SDN improves SDN control by orders-of-magnitude across latency, reliability, and energy-efficiency metrics.

cs.NI↗
arXiv2019

Evolving SDN for Low-Power IoT Networks

Software Defined Networking (SDN) offers a flexible and scalable architecture that abstracts decision making away from individual devices and provides a programmable network platform. However, implementing a centralized SDN architecture within the constraints of a low-power wireless network faces considerable challenges. Not only is controller traffic subject to jitter due to unreliable links and network contention, but the overhead generated by SDN can severely affect the performance of other traffic. This paper addresses the challenge of bringing high-overhead SDN architecture to IEEE 802.15.4 networks. We explore how traditional SDN needs to evolve in order to overcome the constraints of low-power wireless networks, and discuss protocol and architectural optimizations necessary to reduce SDN control overhead - the main barrier to successful implementation. We argue that interoperability with the existing protocol stack is necessary to provide a platform for controller discovery and coexistence with legacy networks. We consequently introduce μSDN, a lightweight SDN framework for Contiki, with both IPv6 and underlying routing protocol interoperability, as well as optimizing a number of elements within the SDN architecture to reduce control overhead to practical levels. We evaluate μSDN in terms of latency, energy, and packet delivery. Through this evaluation we show how the cost of SDN control overhead (both bootstrapping and management) can be reduced to a point where comparable performance and scalability is achieved against an IEEE 802.15.4-2012 RPL-based network. Additionally, we demonstrate μSDN through simulation: providing a use-case where the SDN configurability can be used to provide Quality of Service (QoS) for critical network flows experiencing interference, and we achieve considerable reductions in delay and jitter in comparison to a scenario without SDN.

cs.NI↗
arXiv2019

Isolating SDN Control Traffic with Layer-2 Slicing in 6TiSCH Industrial IoT Networks

Recent standardization efforts in IEEE 802.15.4-2015 Time Scheduled Channel Hopping (TSCH) and the IETF 6TiSCH Working Group (WG), aim to provide deterministic communications and efficient allocation of resources across constrained Internet of Things (IoT) networks, particularly in Industrial IoT (IIoT) scenarios. Within 6TiSCH, Software Defined Networking (SDN) has been identified as means of providing centralized control in a number of key situations. However, implementing a centralized SDN architecture in a Low Power and Lossy Network (LLN) faces considerable challenges: not only is controller traffic subject to jitter due to unreliable links and network contention, but the overhead generated by SDN can severely affect the performance of other traffic. This paper proposes using 6TiSCH tracks, a Layer-2 slicing mechanism for creating dedicated forwarding paths across TSCH networks, in order to isolate the SDN control overhead. Not only does this prevent control traffic from affecting the performance of other data flows, but the properties of 6TiSCH tracks allows deterministic, low-latency SDN controller communication. Using our own lightweight SDN implementation for Contiki OS, we firstly demonstrate the effect of SDN control traffic on application data flows across a 6TiSCH network. We then show that by slicing the network through the allocation of dedicated resources along a SDN control path, tracks provide an effective means of mitigating the cost of SDN control overhead in IEEE 802.15.4-2015 TSCH networks.

cs.NI↗