arXiv subjects

Dario Rossi

Publications and source records attributed to Dario Rossi.

50 records · Page 3Linked to original sources

arXiv2021

A First Look at Class Incremental Learning in Deep Learning Mobile Traffic Classification

The recent popularity growth of Deep Learning (DL) re-ignited the interest towards traffic classification, with several studies demonstrating the accuracy of DL-based classifiers to identify Internet applications' traffic. Even with the aid of hardware accelerators (GPUs, TPUs), DL model training remains expensive, and limits the ability to operate frequent model updates necessary to fit to the ever evolving nature of Internet traffic, and mobile traffic in particular. To address this pain point, in this work we explore Incremental Learning (IL) techniques to add new classes to models without a full retraining, hence speeding up model's updates cycle. We consider iCarl, a state of the art IL method, and MIRAGE-2019, a public dataset with traffic from 40 Android apps, aiming to understand "if there is a case for incremental learning in traffic classification". By dissecting iCarl internals, we discuss ways to improve its design, contributing a revised version, namely iCarl+. Despite our analysis reveals their infancy, IL techniques are a promising research area on the roadmap towards automated DL-based traffic analysis systems.

cs.NI↗

arXiv2021

FENXI: Deep-learning Traffic Analytics at the Edge

Live traffic analysis at the first aggregation point in the ISP network enables the implementation of complex traffic engineering policies but is limited by the scarce processing capabilities, especially for Deep Learning (DL) based analytics. The introduction of specialized hardware accelerators i.e., Tensor Processing Unit (TPU), offers the opportunity to enhance the processing capabilities of network devices at the edge. Yet, to date, no packet processing pipeline is capable of offering DL-based analysis capabilities in the data-plane, without interfering with network operations. In this paper, we present FENXI, a system to run complex analytics by leveraging TPU. The design of FENXI decouples forwarding operations and traffic analytics which operates at different granularities i.e., packet and flow levels. We conceive two independent modules that asynchronously communicate to exchange network data and analytics results, and design data structures to extract flow level statistics without impacting per-packet processing. We prototyped and evaluated FENXI on general-purpose servers considering both adversarial and realistic network conditions. Our analysis shows that FENXI can sustain 100 Gbps line rate traffic processing requiring only limited resources, while also dynamically adapting to variable network conditions.

cs.NI↗

arXiv2021

A First Look at Anycast CDN Traffic

Anycast routing is an IP solution that allows packets to be routed to the topologically nearest server. Over the last years it has been commonly adopted to manage some services running on top of UDP, e.g., public DNS resolvers, multicast rendez-vous points, etc. However, recently the Internet have witnessed the growth of new Anycast-enabled Content Delivery Networks (A-CDNs) such as CloudFlare and EdgeCast, which provide their web services (i.e., TCP traffic) entirely through anycast. To the best of our knowledge, little is known in the literature about the nature and the dynamic of such traffic. For instance, since anycast depends on the routing, the question is how stable are the paths toward the nearest server. To bring some light on this question, in this work we provide a first look at A-CDNs traffic by combining active and passive measurements. In particular, building upon our previous work, we use active measurements to identify and geolocate A-CDNs caches starting from a large set of IP addresses related to the top-100k Alexa websites. We then look at the traffic of those caches in the wild using a large passive dataset collected from a European ISP. We find that several A-CDN servers are encountered on a daily basis when browsing the Internet. Routes to A-CDN servers are very stable, with few changes that are observed on a monthly-basis (in contrast to more the dynamic traffic policies of traditional CDNs). Overall, A-CDNs are a reality worth further investigations.

cs.NI↗

arXiv2020

Heterogeneous Data-Aware Federated Learning

Federated learning (FL) is an appealing concept to perform distributed training of Neural Networks (NN) while keeping data private. With the industrialization of the FL framework, we identify several problems hampering its successful deployment, such as presence of non i.i.d data, disjoint classes, signal multi-modality across datasets. In this work, we address these problems by proposing a novel method that not only (1) aggregates generic model parameters (e.g. a common set of task generic NN layers) on server (e.g. in traditional FL), but also (2) keeps a set of parameters (e.g, a set of task specific NN layer) specific to each client. We validate our method on the traditionally used public benchmarks (e.g., Femnist) as well as on our proprietary collected dataset (i.e., traffic classification). Results show the benefit of our method, with significant advantage on extreme cases.

cs.LG↗

arXiv2016

Stochastic Dynamic Cache Partitioning for Encrypted Content Delivery

In-network caching is an appealing solution to cope with the increasing bandwidth demand of video, audio and data transfer over the Internet. Nonetheless, an increasing share of content delivery services adopt encryption through HTTPS, which is not compatible with traditional ISP-managed approaches like transparent and proxy caching. This raises the need for solutions involving both Internet Service Providers (ISP) and Content Providers (CP): by design, the solution should preserve business-critical CP information (e.g., content popularity, user preferences) on the one hand, while allowing for a deeper integration of caches in the ISP architecture (e.g., in 5G femto-cells) on the other hand. In this paper we address this issue by considering a content-oblivious ISP-operated cache. The ISP allocates the cache storage to various content providers so as to maximize the bandwidth savings provided by the cache: the main novelty lies in the fact that, to protect business-critical information, ISPs only need to measure the aggregated miss rates of the individual CPs and do not need to be aware of the objects that are requested, as in classic caching. We propose a cache allocation algorithm based on a perturbed stochastic subgradient method, and prove that the algorithm converges close to the allocation that maximizes the overall cache hit rate. We use extensive simulations to validate the algorithm and to assess its convergence rate under stationary and non-stationary content popularity. Our results (i) testify the feasibility of content-oblivious caches and (ii) show that the proposed algorithm can achieve within 10\% from the global optimum in our evaluation.

cs.NI↗

arXiv2014

Cost-aware caching: optimizing cache provisioning and object placement in ICN

Caching is frequently used by Internet Service Providers as a viable technique to reduce the latency perceived by end users, while jointly offloading network traffic. While the cache hit-ratio is generally considered in the literature as the dominant performance metric for such type of systems, in this paper we argue that a critical missing piece has so far been neglected. Adopting a radically different perspective, in this paper we explicitly account for the cost of content retrieval, i.e. the cost associated to the external bandwidth needed by an ISP to retrieve the contents requested by its customers. Interestingly, we discover that classical cache provisioning techniques that maximize cache efficiency (i.e., the hit-ratio), lead to suboptimal solutions with higher overall cost. To show this mismatch, we propose two optimization models that either minimize the overall costs or maximize the hit-ratio, jointly providing cache sizing, object placement and path selection. We formulate a polynomial-time greedy algorithm to solve the two problems and analytically prove its optimality. We provide numerical results and show that significant cost savings are attainable via a cost-aware design.

cs.NI↗

arXiv2013

Modeling the interdependency of low-priority congestion control and active queue management

Recently, a negative interplay has been shown to arise when scheduling/AQM techniques and low-priority congestion control protocols are used together: namely, AQM resets the relative level of priority among congestion control protocols. This work explores this issue by (i) studying a fluid model that describes system dynamics of heterogeneous congestion control protocols competing on a bottleneck link governed by AQM and (ii) proposing a system level solution able to reinstate priorities among protocols.

cs.NI↗

arXiv2011

Experimental Assessment of BitTorrent Completion Time in Heterogeneous TCP/uTP swarms

BitTorrent, one of the most widespread used P2P application for file-sharing, recently got rid of TCP by introducing an application-level congestion control protocol named uTP. The aim of this new protocol is to efficiently use the available link capacity, while minimizing its interference with the rest of user traffic (e.g., Web, VoIP and gaming) sharing the same access bottleneck. In this paper we perform an experimental study of the impact of uTP on the torrent completion time, the metric that better captures the user experience. We run BitTorrent applications in a flash crowd scenario over a dedicated cluster platform, under both homogeneous and heterogeneous swarm population. Experiments show that an all-uTP swarms have shorter torrent download time with respect to all-TCP swarms. Interestingly, at the same time, we observe that even shorter completion times can be achieved under careful mixtures of TCP and uTP traffic.

cs.NI↗

arXiv2011

Adaptive Probabilistic Flooding for Multipath Routing

In this work, we develop a distributed source routing algorithm for topology discovery suitable for ISP transport networks, that is however inspired by opportunistic algorithms used in ad hoc wireless networks. We propose a plug-and-play control plane, able to find multiple paths toward the same destination, and introduce a novel algorithm, called adaptive probabilistic flooding, to achieve this goal. By keeping a small amount of state in routers taking part in the discovery process, our technique significantly limits the amount of control messages exchanged with flooding -- and, at the same time, it only minimally affects the quality of the discovered multiple path with respect to the optimal solution. Simple analytical bounds, confirmed by results gathered with extensive simulation on four realistic topologies, show our approach to be of high practical interest.

cs.NI↗

arXiv2010

Rethinking low extra delay background transport protocols

BitTorrent has recently introduced LEDBAT, a novel application-layer congestion control protocol for data exchange. The protocol design starts from the assumption that network bottlenecks are at the access of the network, and that thus user traffic competes creating self-inducing congestion. To relieve from this phenomenon, LEDBAT is designed to quickly infer that self-induced congestion is approaching (by detecting relative changes of the one-way delay in the transmission path), and to react by reducing the sending rate prior that congestion occurs. Prior work has however shown LEDBAT to be affected by a latecomer advantage, where newly arriving connections can starve already existing flows. In this work, we propose modifications to the congestion window update mechanism of the LEDBAT protocol that aim at solving this issue, guaranteeing thus intra-protocol fairness and efficiency. Closed-form expressions for the stationary throughput and queue occupancy are provided via a fluid model, whose accuracy is confirmed by means of ns2 packet level simulations. Our results show that the proposed change can effective solve the latecomer issue, without affecting the other original LEDBAT goals at the same time.

cs.NI↗

arXiv2010

A Survey of Green Networking Research

Reduction of unnecessary energy consumption is becoming a major concern in wired networking, because of the potential economical benefits and of its expected environmental impact. These issues, usually referred to as "green networking", relate to embedding energy-awareness in the design, in the devices and in the protocols of networks. In this work, we first formulate a more precise definition of the "green" attribute. We furthermore identify a few paradigms that are the key enablers of energy-aware networking research. We then overview the current state of the art and provide a taxonomy of the relevant work, with a special focus on wired networking. At a high level, we identify four branches of green networking research that stem from different observations on the root causes of energy waste, namely (i) Adaptive Link Rate, (ii) Interface proxying, (iii) Energy-aware infrastructures and (iv) Energy-aware applications. In this work, we do not only explore specific proposals pertaining to each of the above branches, but also offer a perspective for research.

cs.NI↗

arXiv2010

A hands-on Assessment of Transport Protocols with Lower than Best Effort Priority

Last year, the official BitTorrent client switched to LEDBAT, a new congestion control algorithm targeting a lower-than Best Effort transport service. In this paper, we study this new protocol through packet-level simulations, with a special focus on a performance comparison with other lower-than Best Effort protocols such as TCP-LP and TCP-NICE: our aim is indeed to quantify and relatively weight the level of Low-priority provided by such protocols. Our results show that LEDBAT transport generally achieves the lowest possible level of priority, with the default configurations of TCP-NICE and TCP-LP representing increasing levels of aggressiveness. In addition, we perform a careful sensitivity analysis of LEDBAT performance, by tuning its main parameters in both inter-protocol (against TCP) and intra-protocol (against LEDBAT itself) scenarios. In the inter-protocol case, although in case of misconfiguration LEDBAT competes as aggressively as TCP, however we show that it is not possible to achieve an arbitrary level of low-priority by merely tuning its parameters. In the intra-protocol case, we show that coexistence of legacy flows with slightly dissimilar settings, or experiencing different network conditions, can result in significant unfairness.

cs.NI↗

arXiv2010

The quest for LEDBAT fairness

BitTorrent developers have recently introduced a new application layer congestion control algorithm based on UDP framing at transport layer and currently under definition at the IETF LEDBAT Working Group. LEDBAT is a delay-based protocol which aims at offering a "lower than Best Effort" data transfer service, with a lower priority with respect to elastic TCP and interactive traffic (e.g., VoIP, game). However, in its current specification, LEDBAT is affected by a late-comer advantage: indeed the last flow arriving at the bottleneck is more aggressive due to a wrong estimation of the base delay and finally takes over all resources. In this work, we study several solutions to the late-comer problem by means of packet level simulations and simple analysis: in the investigation process, we individuate the root cause for LEDBAT unfairness and propose effective countermeasures.

cs.NI↗

arXiv2009

News from the Internet congestion control world

A few months ago, the BitTorrent developers announced that the transfer of torrent data in the official client was about to switch to uTP, an application-layer congestion-control protocol using UDP at the transport-layer. This announcement immediately raised an unmotivated buzz about a new, imminent congestion collapse of the whole Internet. Though this reaction was not built on solid technical foundation, nevertheless a legitimate question remains: i.e., whether this novel algorithm is a necessary building block for future Internet applications, or whether it may result in an umpteenth addition to the already well populated world of Internet congestion control algorithms. In this paper, we tackle precisely this issue. The novel protocol is now under discussion at the IETF LEDBAT working group, and has been defined in a draft document in March 2009, whose adoption decision will be taken at the beginning of August 2009. Adhering to the IETF draft definition, we implement the LEDBAT congestion control algorithm and investigate its performance by means of packet-level simulations. Considering a simple bottleneck scenario where LEDBAT competes against either TCP or other LEDBAT flows, we evaluate the fairness of the resource share as well as its efficiency. Our preliminary results show that indeed, there is an undoubted appeal behind the novel application-layer congestion-control protocol. Yet, care must be taken in order to ensure that some important points, such as intra-protocol fairness, are fully clarified in the draft specification -- which we hope that this work can contribute to.

cs.NI↗