Searcharxiv⌕ Search

arXiv subjects

James Roberts

Publications and source records attributed to James Roberts.

17 recordsLinked to original sources
arXiv2022

Size-based scheduling vs fairness for datacenter flows: a queuing perspective

Contrary to the conclusions of a recent body of work where approximate shortest remaining processing time first (SRPT) flow scheduling is advocated for datacenter networks, this paper aims to demonstrate that per-flow fairness remains a preferable objective. We evaluate abstract queuing models by analysis and simulation to illustrate the non-optimality of SRPT under the reasonable assumptions that datacenter flows occur in batches and bursts and not, as usually assumed, individually at the instants of a Poisson process. Results for these models have significant implications for the design of bandwidth sharing strategies for datacenter networks. In particular, we propose a novel "virtual fair scheduling" algorithm that enforces fairness between batches and is arguably simple enough to be implemented in high speed devices.

cs.NI↗
arXiv2021

Accelerating Deep Learning Classification with Error-controlled Approximate-key Caching

While Deep Learning (DL) technologies are a promising tool to solve networking problems that map to classification tasks, their computational complexity is still too high with respect to real-time traffic measurements requirements. To reduce the DL inference cost, we propose a novel caching paradigm, that we named approximate-key caching, which returns approximate results for lookups of selected input based on cached DL inference results. While approximate cache hits alleviate DL inference workload and increase the system throughput, they however introduce an approximation error. As such, we couple approximate-key caching with an error-correction principled algorithm, that we named auto-refresh. We analytically model our caching system performance for classic LRU and ideal caches, we perform a trace-driven evaluation of the expected performance, and we compare the benefits of our proposed approach with the state-of-the-art similarity caching -- testifying the practical interest of our proposal.

cs.NI↗
arXiv2020

Performance Benchmarking of State-of-the-Art Software Switches for NFV

With the ultimate goal of replacing proprietary hardware appliances with Virtual Network Functions (VNFs) implemented in software, Network Function Virtualization (NFV) has been gaining popularity in the past few years. Software switches route traffic between VNFs and physical Network Interface Cards (NICs). It is of paramount importance to compare the performance of different switch designs and architectures. In this paper, we propose a methodology to compare fairly and comprehensively the performance of software switches. We first explore the design spaces of seven state-of-the-art software switches and then compare their performance under four representative test scenarios. Each scenario corresponds to a specific case of routing NFV traffic between NICs and/or VNFs. In our experiments, we evaluate the throughput and latency between VNFs in two of the most popular virtualization environments, namely virtual machines (VMs) and containers. Our experimental results show that no single software switch prevails in all scenarios. It is, therefore, crucial to choose the most suitable solution for the given use case. At the same time, the presented results and analysis provide a deeper insight into the design tradeoffs and identifies potential performance bottlenecks that could inspire new designs.

cs.NI↗
arXiv2019

Cache Subsidies for an Optimal Memory for Bandwidth Tradeoff in the Access Network

While the cost of the access network could be considerably reduced by the use of caching, this is not currently happening because content providers (CPs), who alone have the detailed demand data required for optimal content placement, have no natural incentive to use them to minimize access network operator (ANO) expenditure. We argue that ANOs should therefore provide such an incentive in the form of direct subsidies paid to the CPs in proportion to the realized savings. We apply coalition game theory to design the required subsidy framework and propose a distributed algorithm, based on Lagrangian decomposition, allowing ANOs and CPs to collectively realize the optimal memory for bandwidth tradeoff. The considered access network is a cache hierarchy with per-CP central office caches, accessed by all ANOs, at the apex, and per-ANO dedicated bandwidth and storage resources at the lower levels, including wireless base stations, that must be shared by multiple CPs.

cs.NI↗
arXiv2019

Impact of Traffic Characteristics on Request Aggregation in an NDN Router

The paper revisits the performance evaluation of caching in a Named Data Networking (NDN) router where the content store (CS) is supplemented by a pending interest table (PIT). The PIT aggregates requests for a given content that arrive within the download delay and thus brings an additional reduction in upstream bandwidth usage beyond that due to CS hits. We extend prior work on caching with non-zero download delay (non-ZDD) by proposing a novel mathematical framework that is more easily applicable to general traffic models and by considering alternative cache insertion policies. Specifically we evaluate the use of an LRU filter to improve CS hit rate performance in this non-ZDD context. We also consider the impact of time locality in demand due to finite content lifetimes. The models are used to quantify the impact of the PIT on upstream bandwidth reduction, demonstrating notably that this is significant only for relatively small content catalogues or high average request rate per content. We further explore how the effectiveness of the filter with finite content lifetimes depends on catalogue size and traffic intensity.

cs.NI↗
arXiv2014

Multi-resource fairness: Objectives, algorithms and performance

Designing efficient and fair algorithms for sharing multiple resources between heterogeneous demands is becoming increasingly important. Applications include compute clusters shared by multi-task jobs and routers equipped with middleboxes shared by flows of different types. We show that the currently preferred objective of Dominant Resource Fairness has a significantly less favorable efficiency-fairness tradeoff than alternatives like Proportional Fairness and our proposal, Bottleneck Max Fairness. In addition to other desirable properties, these objectives are equally strategyproof in any realistic scenario with dynamic demand.

cs.NI↗
arXiv2014

Enhanced Cluster Computing Performance Through Proportional Fairness

The performance of cluster computing depends on how concurrent jobs share multiple data center resource types like CPU, RAM and disk storage. Recent research has discussed efficiency and fairness requirements and identified a number of desirable scheduling objectives including so-called dominant resource fairness (DRF). We argue here that proportional fairness (PF), long recognized as a desirable objective in sharing network bandwidth between ongoing flows, is preferable to DRF. The superiority of PF is manifest under the realistic modelling assumption that the population of jobs in progress is a stochastic process. In random traffic the strategy-proof property of DRF proves unimportant while PF is shown by analysis and simulation to offer a significantly better efficiency-fairness tradeoff.

cs.PF↗
arXiv2013

Exploring the Memory-Bandwidth Tradeoff in an Information-Centric Network

An information-centric network should realize significant economies by exploiting a favourable memory-bandwidth tradeoff: it is cheaper to store copies of popular content close to users than to fetch them repeatedly over the Internet. We evaluate this tradeoff for some simple cache network structures under realistic assumptions concerning the size of the content catalogue and its popularity distribution. Derived cost formulas reveal the relative impact of various cost, traffic and capacity parameters, allowing an appraisal of possible future network architectures. Our results suggest it probably makes more sense to envisage the future Internet as a loosely interconnected set of local data centers than a network like today's with routers augmented by limited capacity content stores.

cs.NI↗
arXiv2013

The Dimensional Reduction and Kähler Metric of Forms In Flux and Warping

We present a first-principles derivation of the Kähler metric for axion-like moduli of conformally Calabi-Yau compactifications of IIB string theory with imaginary self-dual 3-form flux at the classical level. We find that the warp factor and flux modify the moduli space metric and therefore Kähler potential even in classical supergravity, with the modifications scaling as (volume)$^{-2/3}$ in the large-volume limit. Our derivation emphasizes the role of constraints from 10D gauge symmetries and highlights metric formality as a geometric property that protects the moduli space of highly supersymmetric toroidal orientifolds. Our results have important quantitative implications for nonperturbative moduli stabilization, phenomenology, and cosmology in flux compactifications.

hep-th↗
arXiv2012

Best-effort networks: modeling and performance analysis via large networks asymptotics

In this paper we introduce a class of Markov models, termed best-effort networks, designed to capture performance indices such as mean transfer times in data networks with best-effort service. We introduce the so-called min bandwidth sharing policy as a conservative approximation to the classical max-min policy. We establish necessary and sufficient ergodicity conditions for best-effort networks under the min policy. We then resort to the mean field technique of statistical physics to analyze network performance deriving fixed point equations for the stationary distribution of large symmetrical best-effort networks. A specific instance of such net- works is the star-shaped network which constitutes a plausible model of a network with an overprovisioned backbone. Numerical and analytical study of the equations allows us to state a number of qualitative conclusions on the impact of traffic parameters (link loads) and topology parameters (route lengths) on mean document transfer time.

math.PR↗
arXiv2012

A versatile and accurate approximation for LRU cache performance

In a 2002 paper, Che and co-authors proposed a simple approach for estimating the hit rates of a cache operating the least recently used (LRU) replacement policy. The approximation proves remarkably accurate and is applicable to quite general distributions of object popularity. This paper provides a mathematical explanation for the success of the approximation, notably in configurations where the intuitive arguments of Che, et al clearly do not apply. The approximation is particularly useful in evaluating the performance of current proposals for an information centric network where other approaches fail due to the very large populations of cacheable objects to be taken into account and to their complex popularity law, resulting from the mix of different content types and the filtering effect induced by the lower layers in a cache hierarchy.

cs.NI↗
arXiv2012

Impact of traffic mix on caching performance in a content-centric network

For a realistic traffic mix, we evaluate the hit rates attained in a two-layer cache hierarchy designed to reduce Internet bandwidth requirements. The model identifies four main types of content, web, file sharing, user generated content and video on demand, distinguished in terms of their traffic shares, their population and object sizes and their popularity distributions. Results demonstrate that caching VoD in access routers offers a highly favorable bandwidth memory tradeoff but that the other types of content would likely be more efficiently handled in very large capacity storage devices in the core. Evaluations are based on a simple approximation for LRU cache performance that proves highly accurate in relevant configurations.

cs.NI↗
arXiv2011

Getting routers out of the core: Building an optical wide area network with "multipaths"

We propose an all-optical networking solution for a wide area network (WAN) based on the notion of multipoint-to-multipoint lightpaths that, for short, we call "multipaths". A multipath concentrates the traffic of a group of source nodes on a wavelength channel using an adapted MAC protocol and multicasts this traffic to a group of destination nodes that extract their own data from the confluent stream. The proposed network can be built using existing components and appears less complex and more efficient in terms of energy consumption than alternatives like OPS and OBS. The paper presents the multipath architecture and compares its energy consumption to that of a classical router-based ISP network. A flow-aware dynamic bandwidth allocation algorithm is proposed and shown to have excellent performance in terms of throughput and delay.

cs.NI↗
arXiv2011

A Flow-aware MAC Protocol for a Passive Optical Metropolitan Area Network

The paper introduces an original MAC protocol for a passive optical metropolitan area network using time-domain wavelength interleaved networking (TWIN)% as proposed recently by Bell Labs . Optical channels are shared under the distributed control of destinations using a packet-based polling algorithm. This MAC is inspired more by EPON dynamic bandwidth allocation than the slotted, GPON-like access control generally envisaged for TWIN. Management of source-destination traffic streams is flow-aware with the size of allocated time slices being proportional to the number of active flows. This emulates a network-wide, distributed fair queuing scheduler, bringing the well-known implicit service differentiation and robustness advantages of this mechanism to the metro area network. The paper presents a comprehensive performance evaluation based on analytical modelling supported by simulations. The proposed MAC is shown to have excellent performance in terms of both traffic capacity and packet latency.

cs.NI↗
arXiv2010

Traffic Capacity of Large WDM Passive Optical Networks

As passive optical networks (PON) are increasingly deployed to provide high speed Internet access, it is important to understand their fundamental traffic capacity limits. The paper discusses performance models applicable to wavelength division multiplexing (WDM) EPONs and GPONs under the assumption that users access the fibre via optical network units equipped with tunable transmitters. The considered stochastic models are based on multiserver polling systems for which explicit analytical results are not known. A large system asymptotic, mean-field approximation, is used to derive closed form solutions of these complex systems. Convergence of the mean field dynamics is proved in the case of a simple network configuration. Simulation results show that, for a realistic sized PON, the mean field approximation is accurate.

cs.NI↗
arXiv2010

Upstream traffic capacity of a WDM EPON under online GATE-driven scheduling

Passive optical networks are increasingly used for access to the Internet and it is important to understand the performance of future long-reach, multi-channel variants. In this paper we discuss requirements on the dynamic bandwidth allocation (DBA) algorithm used to manage the upstream resource in a WDM EPON and propose a simple novel DBA algorithm that is considerably more efficient than classical approaches. We demonstrate that the algorithm emulates a multi-server polling system and derive capacity formulas that are valid for general traffic processes. We evaluate delay performance by simulation demonstrating the superiority of the proposed scheduler. The proposed scheduler offers considerable flexibility and is particularly efficient in long-reach access networks where propagation times are high.

cs.NI↗