IMDEA Networks Institute Digital Repository
Not a member yet
1915 research outputs found
Sort by
The Rumble in the Millimeter Wave Jungle: Obstructions Vs RIS
No full textReconfigurable intelligent surfaces (RISs) have emerged as a key technology for future communication systems. RISs are arrays of tunable reflecting elements that provide controllable propagation channels by smartly shaping incident electromagnetic (EM) waves. Analysis and improvement of RIS-aided systems require the definition of accurate path loss models that consider environmental effects often encountered in practical applications. In this paper, we derive a path loss model for RIS-assisted communications to account for the attenuation induced by the transmission medium and randomly located obstructions. More precisely, this study focuses on assessing the impact caused by Poisson-located obstructing objects on RIS-assisted millimeter wave links. To this end, we evaluate the outage probability yielded by RIS-aided systems in indoor environments with antenna beam-steering and random obstructions. We obtain extensive simulation results to assess the impact of RIS considering different parameters, such as the minimum signal-to-noise ratio (SNR) necessary for successful reception, the operating frequency, the density of the Poisson process used for object placement, and the object size.Ministry of Economic Affairs and Digital Transformation, European Union NextGeneration-EUMario Gerla Best Paper AwardTRUEinpres
Evaluating the Impact of Flow Length on the Performance of In-Switch Inference Solutions
Get PDFWorkshop name: The 11th International Workshop on Computer and Networking Experimental Research using Testbeds (CNERT)As modern networks evolve into complex systems to support next-generation applications with strict latency requirements, in-switch machine learning (ML) has emerged as a candidate technology for minimizing ML inference latency. Multiple solutions, mostly based on Decision Tree (DT) and Random Forest (RF) models, have been proposed in that regard for inference at packet level (PL) or flow level (FL) or simultaneously at PL and FL. Such heterogeneity in the inference target leads to the use of varying performance metrics for evaluating the solutions, rendering a fair comparison between them difficult. In this paper, we perform a comprehensive evaluation of 5 leading solutions for DT/RF-based in-switch inference. We replicate and deploy the solutions into a real-world testbed with Intel Tofino switches, and run experiments with measurement data from 4 datasets. We then evaluate their performance using (i) the original metric used in the solution’s paper, and (ii) a novel FL metric which evaluates every solution at FL. This FL metric enables us to delve into an extensive analysis of how the solutions perform on flows of different lengths in diverse use cases. Results show that while some solutions perform similarly across use cases and flow sizes, others show inconsistent behaviours that we discuss.Spanish Ministry of Economic Affairs and Digital Transformation and the European Union-NextGenerationEU through the UNICO 5G I+D project no.TSI-063000-2021-52 “AEON-ZERO"Spanish Ministry of Science and Innovation through grant no. PID2021-128250NB-I00 “bRAIN”CHIST-ERA grant no. CHIST-ERA-20-SICT- 001 “ECOMOME”, via grant PCI2022-133013 of Agencia Estatal de InvestigaciónEuropean Union’s Horizon Europe research and innovation programme under grant agreement no. 101017109 "DAEMON"TRUEinpres
An Evaluation of RAN Sustainability Strategies in Production Networks
Get PDFReducing energy consumption is a primary goal for the mobile telecommunication industry, with strong environmental and economic implications. The main target for savings is the Radio Access Network (RAN), which is responsible for more than 70% of the total energy costs incurred by operators. Lowering energy costs at the RAN is possible by reducing the number of active carriers at off-peak locations and times where the demand can be served with a lower capacity than deployed. While the scientific community has been proposing a plethora of complex solutions to switch-off underutilized carriers, production networks largely rely nowadays on threshold-based strategies that run at individual RAN equipment and are typically enabled only overnight. Moreover, there are no real-world evaluations of the effectiveness of carrier switch-off approaches in reducing energy consumption or their impact on the end users. In this paper, we benchmark five fixed threshold-based cell sleep policies deployed in a production network serving large geographical regions. The study provides unprecedented insights on industry-grade RAN sustainability at scale, in terms of actual energy savings and trade-offs with user experience. Our insights suggest that the capability of the tested policies in reducing the energy costs hits a clear barrier if no degradation is admissible for any user, and provides a strong empirical basis in support of more flexible approaches to save energy at the RAN.Comunidad de MadridEuropean UnionSpanish Ministry of Economic Affairs and Digital TransformationEuropean Union – NextGeneration EUTRUEinpres
YinYangRAN: Resource Multiplexing in GPU-Accelerated Virtualized RANs
No full textRAN virtualization is revolutionizing the telco industry, enabling 5G Distributed Units to run using general-purpose platforms equipped with Hardware Accelerators (HAs). Recently, GPUs have been proposed as HAs, hinging on their unique capability to execute 5G PHY operations efficiently while also processing Machine Learning (ML) workloads. While this ambivalence makes GPUs attractive for cost-effective deployments, we experimentally demonstrate that multiplexing 5G and ML workloads in GPUs is in fact challenging, and that using conventional GPU-sharing methods can severely disrupt 5G operations. We then introduce YinYangRAN, an innovative O-RAN-compliant solution that supervises GPU-based HAs so as to ensure reliability in the 5G processing pipeline while maximizing the throughput of concurrent ML services. YinYangRAN performs GPU resource allocation decisions via a computationally-efficient approximate dynamic programming technique, which is informed by a neural network trained on real-world measurements. Using workloads collected in real RANs, we demonstrate that YinYangRAN can achieve over 50\% higher 5G processing reliability than conventional GPU sharing models with minimal impact on co-located ML workloads. To our knowledge, this is the first work identifying and addressing the complex problem of HA management in emerging GPU-accelerated vRANs, and represents a promising step towards multiplexing PHY and ML workloads in mobile networks.Spanish Ministry of Economic Affairs and Digital Transformation and the European Union-NextGenerationEU through the UNICO 5G I+D project no.TSI-063000-2021-52 “AEON-ZERO"European Commission through Grant No. SNS-JU-101097083 (BeGREEN), 101139270 (ORIGAMI), and 101017109 (DAEMON)TRUEinpres
Impact of Public Protests on Mobile Networks
Get PDFThe roll out of 5G, coupled with the traffic monitoring capabilities of modern industry-grade networks, offers an unprecedented opportunity to closely observe the impact that the introduction of a new major wireless technology has on the end users.
In this paper, we seize such a unique chance, and carry out a first-of-its-kind in-depth analysis of 5G adoption along spatial, temporal and service dimensions. Leveraging massive measurement data about application-level demands collected in a nationwide 4G/5G network, we characterize the impact of the new technology on when, where and how mobile subscribers consume 5G traffic both in aggregate and for individual types of services.
This lets us unveil the overall incidence of 5G in the total mobile network traffic, its spatial and temporal fluctuations, its effect on the way 5G services are consumed, the way individual services and geographical locations contribute to fluctuations in the 5G demand, as well as surprising connections between socioeconomic status of local populations and the way the 5G technology is presently consumed.TRUEinpres
Effectiveness of Distributed Stateless Network Server Selection under Strict Latency Constraints
Get PDFWe consider a set of network users (nodes), each generating latency-constrained service requests corresponding to the execution of computational tasks on servers positioned either within a cloud infrastructure or at the network edge. Within this framework, we systematically assess the efficacy of a distributed stateless server selection strategy, strategically performed by individual nodes. Leveraging principles from game theory, our study allows for a comparative analysis between the optimality achieved through globally orchestrated stateless allocation and a decentralized stateless server selection mechanism driven by the self-interested objectives of individual nodes.
Our emphasis on stateless server allocation, rooted in a probabilistic selection framework between edge and cloud servers, stems from prior empirical revelations demonstrating the advantageous outcomes of determining the optimal distribution of edge and cloud tasks based on static network characteristics. Importantly, this determination occurs irrespective of the real-time network state.
The suboptimal nature of the selfish allocation is quantified by the so called “price of anarchy,” a metric shown to approximate unity closely. This observation substantiates the justification for a distributed strategic implementation of stateless policies. This elucidation serves as a pivotal guide for crafting algorithms governing server selection, providing a quantitative validation of the efficacy inherent in distributed self-interested approaches.TRUEpu
Characterizing Large-Scale Mobile Traffic Measurements for Urban, Social and Networks Sciences
Get PDFOver the last few decades, it is difficult to pinpoint a technological advancement that
shifted the daily life of the world’s population in a more disrupting way than mobile phones
and their applications. Their ubiquitousness has reshaped global behaviors and routines,
transforming portable devices into the essential and on-the-go personal computer. Mobile
phones enable communication, information access, and entertainment with little to no
location constraints, thanks to their connectivity to the internet through a pervasive radio
access network infrastructure. Of course, this seamless mobile access was not always a
given, and decades of research, development and technology integration were needed to
reach today’s high-capacity support for broadband and low-latency mobile services.
From the first generations of mobile networks offering only on-the-go voice and
text to the current fourth and fifth generations supporting high-resolution on-demand
video streaming and low-latency cloud applications, every new release contributed
to transforming mobile phones into essential items.
With the rise in popularity of
mobile applications in smartphones, any company or developer could release their own
application, giving access to their product to consumers anywhere. Advancements in
mobile networks meant an increase in data transfer capacities, leading users to be more
comfortable utilizing their smartphones for tasks anywhere and at any time. The sucess of
mobile technologies also signifies that the patterns of usage captured by mobile operators
reflect in a rich and detailed way the endeavors of their vast user population.
Due to this reason, the data collected in operational mobile networks has today
become a primary source of information for research in networking and beyond. Early
research utilized analysis of mobile network traffic as feedback for the mobile operator
itself, as a way to understand the spatiotemporal dynamics of the operational demands
in the network, and how this could be leveraged to improve network deployments and
operation according to consumption patterns. A second and broader direction lies in
interdisciplinary research, seeking to explore how these measurements could be used to
understand populations and urban environment dynamics.
This drives the need of research oriented towards networks data science: the study
of tools and methodologies capable of handling large-scale measurements, asserting the
quality and precision of the collected data in reflecting reality, as well as developing tools capable of extracting insights and making the vastness of collected information useful for
analysis. This thesis is a step in the direction of establishing said tools and methodologies,
as well as showcasing several potential directions that mobile network measurement can
support in interdisciplinary research domains.
The first part of this thesis presents a full contextualization of networks data science,
expanding on the problem of the ever-growing scale of collected sets and presenting the
many different fields that have been explored over the years, from classical network
engineering applications to the study of populations, epidemics, socioeconomic and
people’s movement and transportation across cities.
These studies are not possible
without a well-established routine of data collection and processing, which is also discussed
in the first part of this thesis.
The second part of the manuscript presents the original contributions provided by
the thesis. Four chapters explore different directions where the collected data can be
leveraged to derive new insights. First, an overview of the adoption of new technologies
provided by the mobile network operator, with new findings into how changes in their
traffic patterns may happen according to these new capabilities. Second, an exploration of
how mobile consumption patterns and demands can be utilized to better understand the
space within cities, with new methodologies presenting how both city-wide and location-
specific insights can be gained just by looking at the traffic being consumed within
base stations of the network. Third, a look into how special events may impact mobile
networks, as such occurrences affect directly how users interact with their smartphones. It
becomes important for the network operator to be able to extract insights and understand
how these variations in traffic demand across time, space, and applications affect the
functioning of the network, as well as these insights can be used by lawmakers to
understand how these events affect populations. Lastly, a study characterizing session-
level measurements to derive insights used to generate synthetic data sets containing new
dynamics, generating simple models that can be utilized by anyone interested in research
within mobile networks to test and validate their data-driven solutions.
In summary, the age of pervasive digital services leaves researchers with oceans of data
and information to be explored in many different areas, with mobile networks shaping
into a major source of rich information to guide innovation in both cutting edge research
and technology development. This thesis guides the reader through the current state of
affairs, showcasing the current opportunities opened by mobile network data and also
presenting potential future directions that can be pursed in the following years.Telematics EngineeringUniversidad Carlos III de Madrid, Spai
Jewel: Resource-Efficient Joint Packet and Flow Level Inference in Programmable Switches
Get PDFEmbedding machine learning (ML) models in programmable switches realizes the vision of high-throughput and low-latency inference at line rate. Recent works have made breakthroughs in embedding Random Forest (RF) models in switches for either packet-level inference or flow-level inference. The former relies on simple features from packet headers that are simple to implement but limit accuracy in challenging use cases; the latter exploits richer flow features to improve accuracy, but leaves early packets in each flow unclassified. We propose Jewel, an in-switch ML model based on a fully joint packet- and flow-level design, which takes the best of both worlds by classifying early flow packets individually and shifting to flow-level inference when possible. Our proposal involves (i) a single RF model trained to classify both packets and flows, and (ii) hardware-aware model selection and training techniques for resource footprint minimization. We implement Jewel in P4 and deploy it in a testbed with Intel Tofino switches, where we run extensive experiments with a variety of real-world use cases. Results reveal how our solution outperforms four state-of-the-art benchmarks, with accuracy gains in the 2.0%–5.3% range.CHIST-ERA grant no. CHIST-ERA-20-SICT- 001 “ECOMOME”, via grant PCI2022-133013 of Agencia Estatal de InvestigaciónSpanish Ministry of Science and Innovation through grant no. PID2021-128250NB-I00 “bRAIN”Spanish Ministry of Economic Affairs and Digital Transformation and the European Union-NextGenerationEU through the UNICO 5G I+D project no.TSI-063000-2021-52 “AEON-ZERO"TRUEinpres
Strengthening Privacy in Robust Federated Learning through Secure Aggregation
Get PDFFederated Learning (FL) has evolved into a pivotal paradigm for collaborative machine learning, enabling a centralised server to compute a global model by aggregating the local models trained by clients. However, the distributed nature of FL renders it susceptible to poisoning attacks that exploit its linear aggregation rule called FEDAVG. To address this vulnerability, FEDQV has been recently introduced as a superior alternative to FEDAVG, specifically designed to mitigate poisoning attacks by taxing more than linearly deviating clients. Nevertheless, FEDQV remains exposed to privacy attacks that aim to infer private
information from clients’ local models. To counteract such privacy threats, a well-known approach is to use a Secure Aggregation (SA) protocol to ensure that the server is unable to inspect individual trained models as it aggregates them. In this work, we show how to implement SA on top of FEDQV in order to
address both poisoning and privacy attacks. We mount several privacy attacks against FEDQV and demonstrate the effectiveness of SA in countering them.Ministry of Economic Affairs and Digital Transformation, European Union NextGeneration-EUTRUEpu
CloudRIC: Open Radio Access Network (O-RAN) Virtualization with Shared Heterogeneous Computing
Get PDFOpen and virtualized Radio Access Networks (vRANs) are breeding a new market with unprecedented opportunities. However, carrier-grade vRANs today are expensive and energyhungry, as they rely on hardware accelerators (HAs) that are dedicated to individual distributed units (DUs). In this paper, we argue that sharing pools of heterogeneous processors among DUs leads to more cost- and energy-efficient vRANs. We then design CloudRIC, a system that, powered by lightweight data-driven models, meets specific reliability targets while (��) coordinating access between DUs and heterogeneous computing infrastructure; and (����) assisting DUs with compute-aware radio scheduling procedures. Experiments on a GPU-accelerated O-Cloud show that CloudRIC can achieve, respectively, 3x and 15x mean gains in energy and cost-efficiency under real RAN workloads while ensuring 99.999% reliability even in dense scenarios.Spanish Ministry of Economic Affairs and Digital Transformation and the European Union-NextGenerationEU through the UNICO 5G I+D project no.TSI-063000-2021 "OPEN6G"Spanish Ministry of Economic Affairs and Digital Transformation and the European Union-NextGenerationEU through the UNICO 5G I+D project no. 022/0005395 "CLARION"TRUEpu