INFOCOM 2021 Papers

NFV ecosystem is flourishing and more and more NF platforms appear, but this makes NF vendors difficult to deliver NFs rapidly to diverse platforms. We propose an NF development framework named NFD for cross-platform NF development. NFD's main idea is to decouple the functional logic from the platform logic -it provides a platform-independent language to program NFs' behavior models, and a compiler with interfaces to develop platform-specific plugins. By enabling a plugin on the compiler, various NF models would be compiled to executables integrated with the target platform. We prototype NFD, build 14 NFs, and support 6 platforms (standard Linux, OpenNetVM, GPU, SGX, DPDK, OpenNF). Our evaluation shows that NFD can save development workload for cross-platform NFs and output valid and performant NFs.

Acquiring the geographical distribution of neighbors can support more adaptive media access control (MAC) protocols and other safety applications in Vehicular ad hoc network (VANETs). However, it is very challenging for each vehicle to estimate its own neighbor distribution in a fully distributed setting. In this paper, we propose an online distributed neighbor distribution estimation scheme, called PeerProbe, in which vehicles collaborate with each other to probe their own neighborhood via simultaneous symbol-level wireless communication. An adaptive compressive sensing algorithm is developed to recover a neighbor distribution based on a small number of random probes with non-negligible noise. Moreover, the needed number of probes adapts to the sparseness of the distribution. We conduct extensive simulations and the results demonstrate that PeerProbe is lightweight and can accurately recover highly dynamic neighbor distributions in critical channel conditions.

The number of requests to transfer large files is increasing rapidly in web server and remote-storage scenarios, and this increase requires a higher processing capacity from the network stack. However, to fully decouple from applications, many latest userspace network stacks, such as VPP (vector packet processing) and snap, adopt a shared-memory-based solution to communicate with upper applications. During this communication, the application or network stack needs to copy data to or from shared memory queues. In our verification experiment, these multiple copy operations incur more than 50% CPU consumption and severe performance degradation when the transferred file is larger than 32 KB. This paper adopts a hardware-accelerated solution and proposes HAVS which integrates Intel I/O Acceleration Technology into the VPP network stack to achieve high-performance memory copy offloading. An asynchronous copy architecture is introduced in HAVS to free up CPU resources. Moreover, an abstract memcpy accelerator layer is constructed in HAVS to ease the use of different types of hardware accelerators and sustain high availability with a fault-tolerance mechanism. The comprehensive evaluation shows that HAVS can provide an average 50%-60% throughput improvement over the original VPP stack when accelerating the nginx and SPDK iSCSI target application.

Data generated at the network edge can be processed locally by leveraging the paradigm of edge computing. To fully utilize the widely distributed data, we concentrate on a wireless edge computing system that conducts model training using decentralized peer-to-peer (P2P) methods. However, there are two major challenges on the way towards efficient P2P model training: limited resources (e.g., network bandwidth and battery life of mobile edge devices) and time-varying network connectivity due to device mobility or wireless channel dynamics, which have received less attention in recent years. To address these two challenges, this paper adaptively constructs a dynamic and efficient P2P topology, where model aggregation occurs at the edge devices. In a nutshell, we first formulate the topology construction for P2P learning (TCPL) problem with resource constraints as an integer programming problem. Then a learning-driven method is proposed to adaptively construct a topology at each training epoch. We further give the convergence analysis on training machine learning models even with non-convex loss functions. Extensive simulation results show that our proposed method can improve the model training efficiency by about 11% with resource constraints and reduce the communication cost by about 30% under the same accuracy requirement compared to the benchmarks.

Accurate identification of the signal type in shared-spectrum networks is critical for efficient resource allocation and fair coexistence. It can be used for scheduling transmission opportunities to avoid collisions and improve system throughput, especially when the environment changes rapidly. In this paper, we develop deep neural networks (DNNs) to detect coexisting signal types based on In-phase/Quadrature (I/Q) samples without decoding them. By using segments of the samples of the received signal as input, a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN) are combined and trained using categorical cross-entropy (CE) optimization. Classification results for coexisting Wi-Fi, LTE LAA, and 5G NR-U signals in the 5-6 GHz unlicensed band show high accuracy of the proposed design. We then exploit spectrum analysis of the I/Q sequences to further improve the classification accuracy. By applying Short-time Fourier Transform (STFT), additional information in the frequency domain can be presented as a spectrogram. Accordingly, we enlarge the input size of the DNN. To verify the effectiveness of the proposed detection framework, we conduct over-the-air (OTA) experiments using USRP radios. The proposed approach can achieve accurate classification in both simulations and hardware experiments.

Fairness is a crucial design objective in virtually all network optimization problems, where limited system resources are shared by multiple agents. Recently, reinforcement learning has been successfully applied to autonomous online decision making in many network design and optimization problems. However, most of them try to maximize the long-term (discounted) reward of all agents, without taking fairness into account. In this paper, we propose a family of algorithms that bring fairness to actor-critic reinforcement learning for optimizing general fairness utility functions. In particular, we present a novel method for adjusting the rewards in standard reinforcement learning by a multiplicative weight depending on both the shape of fairness utility and some statistics of past rewards. It is shown that for proper choice of the adjusted rewards, a policy gradient update converges to at least a stationary point of general αfairness utility optimization. It inspires the design of fairness optimization algorithms in actor-critic reinforcement learning. Evaluations show that the proposed algorithm can be easily deployed in real-world network optimization problems, such as wireless scheduling and video QoE optimization, and can significantly improve the fairness utility value over previous heuristics and learning algorithms.

Container-based microservice has emerged as a promising technique in promoting edge computing elasticity. At the runtime, microservices, encapsulated in form of container images, need to be frequently downloaded from remote registries to local edge servers, which may incur significant overhead in terms of excessive download traffic and large local storage. Given the limited resources at the edge, it is of critical importance to minimize such overhead in order to enhance microservice offerings. A distinctive feature in container-based microservice, which has not been exploited, is that microservice images are in layered structure and common layers can be shared by co-located microservices. In this paper, we study a layer aware micro-service placement and request scheduling at the edge. Intuitively, throughput and number of hosted microservices can be significantly increased by layer sharing between co-located images. We formulate this into an optimization problem with approximate submodularity, and prove this to be NP-hard. We design an iterative greedy algorithm with guaranteed approximation ratio. Extensive experiments validate the efficiency of our method, and the results demonstrate that the number of placed microservices can be increased by 27.61% and the microservice throughput can be improved by 73.13%, respectively, in comparison with the state-of-the-art microservice placement strategy.

For smart grid services, accurate individual load forecasting is an essential element. When training individual forecasting models for multi-customers, discrepancies in data distribution among customers should be considered; there are two simple ways to build the models considering multi-customers: constructing each model independently or training as one model encompassing multi-customers. The independent approach shows higher accuracy than the latter. However, it deploys copious models, causing resource/management inefficiency; the latter is the opposite. A compromise between these two could be clustering-based forecasting. However, the previous studies are limited in applying to individual forecasting in that they focus on aggregated load and do not consider concept drift, which degrades accuracy over time. Therefore, we propose a distribution-aware temporal pooling framework that is enhanced clustering-based forecasting. For the clustering, we propose Variational Recurrent Deep Embedding (VaRDE) working in a distribution-aware manner, so it is suitable to process individual load. It allocates clusters to customers every time, so the clusters, where customers are assigned, are dynamically changed to resolve distribution change. We conducted experiments with real data for evaluation, and the result showed better performance than previous studies, especially with a few models even for unseen data, leading to high scalability.

Operating federated learning optimally over distributed cloud-edge networks is a non-trivial task, which requires to manage data transference from user devices to edges, resource provisioning at edges, and federated learning between edges and the cloud. We formulate a non-linear mixed integer program, minimizing the long-term cumulative cost of such a federated learning system while guaranteeing the desired convergence of the machine learning models being trained. We then design a set of novel polynomial-time online algorithms to make adaptive decisions by solving continuous solutions and converting them to integers to control the system on the fly, based only on the predicted inputs about the dynamic and uncertain cloud-edge environments via online learning. We rigorously prove the competitive ratio, capturing the multiplicative gap between our approach using predicted inputs and the offline optimum using actual inputs. Extensive evaluations with real-world training datasets and system parameters confirm the empirical superiority of our approach over multiple state-of-the-art algorithms.

Indoor localization serves an important role in various scenarios such as navigation in shopping malls or hospitals. However, the existing technology is usually based on additional deployment and the signals suffer from strong environmental interference in the complex indoor environment. In this paper, we propose video-based indoor localization with text information (i.e. "VideoLoc") without the deployment of additional equipment. Videos taken by the phone carriers cover more critical information (e.g. logos in malls), while a single photo may fail to capture it. To reduce redundant information in the video, we propose key-frame selection based on deep learning model and clustering algorithm. Video frames are characterized with deep visual descriptors and the clustering algorithm efficiently clusters these descriptors into a set of non-overlapping snippets. We select keyframes from these non-overlapping snippets in terms of the cluster centroid that represents each snippet. Then, we propose text detection and recognition with the perspective transformation to make full use of stable and discriminative text information (e.g. logos or room numbers) in keyframes for localization. Finally, we obtain the location of the phone carrier via the triangulation algorithm. The experimental results show that VideoLoc achieves high precision of localization and is robust to dynamic environments.

Global IPv6 scanning has always been a challenge for researchers because of the limited network speed and computational power. Target generation algorithms are recently proposed to overcome the problem for Internet assessments by predicting a candidate set to scan. However, IPv6 custom address configuration emerges diverse addressing patterns discouraging algorithmic inference. Widespread IPv6 alias could also mislead the algorithm to discover aliased regions rather than valid host targets. In this paper, we introduce 6GAN, a novel architecture built with Generative Adversarial Net (GAN) and reinforcement learning for multi-pattern target generation. 6GAN forces multiple generators to train with a multi-class discriminator and an alias detector to generate non-aliased active targets with different addressing pattern types. The rewards from the discriminator and the alias detector help supervise the address sequence decision-making process. After adversarial training, 6GAN's generators could keep a strong imitating ability for each pattern and 6GAN's discriminator obtains outstanding pattern discrimination ability with a 0.966 accuracy. Experiments indicate that our work outperformed the state-of-the-art target generation algorithms by reaching a higher-quality candidate set.

Mobile cloud computing enables compute-limited mobile devices to perform real-time intensive computations such as speech recognition or object detection by leveraging powerful cloud servers. An important problem in large-scale mobile cloud computing is computational offloading where each mobile device decides when and how much computation should be uploaded to cloud servers by considering the local processing delay and the cost of using cloud servers. In this paper, we develop a distributed threshold-based offloading algorithm where it uploads an incoming computing task to cloud servers if the number of tasks queued at the device reaches the threshold, and processes it locally otherwise. The threshold is updated iteratively based on the computational load and the cost of using cloud servers. We formulate the problem as a symmetric game, and characterize the sufficient and necessary conditions for the existence and uniqueness of the Nash Equilibrium (NE) assuming exponential service times. Then, we show the convergence of our proposed distributed algorithm to the NE when the NE exists. Finally, we perform extensive simulations to validate our theoretical findings and demonstrate the efficiency of our proposed distributed algorithm under various practical scenarios such as general service times, imperfect server utilization estimation, and asynchronous threshold updates.

Safety-critical networks often have stringent real-time requirements; they must also be resilient to failures. In this paper, we propose the RealFlow framework that uses commodity software-defined networks (SDNs) to realize networks with end-to-end timing guarantees, while also: (a) increasing resiliency against link/switch failures and (b) increasing network utilization. The use of SDNs in this space also improves the management capabilities of the system due to the global visibility into the network. RealFlow is implemented as a northbound SDN controller application compatible with standard OpenFlow protocols with little to no runtime overheads. We demonstrate feasibility on a real hardware testbed (Pica8 SDN switches+Raspberry Pi endhosts) and a practical avionics case study. Our evaluations show that RealFlow can accommodate 63% more network flows with safety-critical guarantees when compared to current designs and up to 18% when link resiliency (via backup paths) is also considered.

The mobile network operators are upgrading their network facilities and shifting to the 5G era at an unprecedented pace. The huge operating expense (OPEX), mainly the energy consumption cost, has become the major concern of the operators. In this work, we investigate the energy cost-saving potential by transforming the backup batteries of base stations (BSs) to a distributed battery energy storage system (BESS). Specifically, to minimize the total energy cost, we model the distributed BESS discharge/charge scheduling as an optimization problem by incorporating comprehensive practical considerations. Then, considering the dynamic BS power demands in practice, we propose a deep reinforcement learning (DRL) based approach to make BESS scheduling decisions in real-time. The experiments using real-world BS deployment and traffic load data demonstrate that with our DRL-based BESS scheduling, the peak power demand charge of BSs can be reduced by up to 26.59%, and the yearly OPEX saving for 2,282 5G BSs could reach up to US$185,000.

Many online platforms are providing valuable real-time contents (e.g., traffic) by continuously acquiring the status of different Points of Interest (PoIs). In status acquisition, it is challenging to determine how frequently a PoI should upload its status to a platform, since they are self-interested with private and possibly time-varying preferences. This paper considers a general multi-period status acquisition system, aiming to maximize the aggregate social welfare and ensure the platform freshness. The freshness is measured by a metric termed age of information. For this goal, we devise a long-term decomposition (LtD) mechanism to resolve the time-varying information asymmetry. The key idea is to construct a virtual social welfare that only depends on the current private information, and then decompose the per-period operation into multiple distributed bidding problems for the PoIs and platforms. The LtD mechanism enables the platforms to achieve a tunable trade-off between payoff maximization and freshness conditions. Moreover, the LtD mechanism retains the same social performance compared to the benchmark with symmetric information and asymptotically ensures the platform freshness conditions. Numerical results based on real-world data show that when the platforms pay more attention to payoff maximization, each PoI still obtains a non-negative payoff in the long-term.

Radio Access Network Virtualization (vRAN) will spearhead the quest towards supple radio stacks that adapt to heterogeneous infrastructure: from energy-constrained platforms deploying cells-on-wheels (e.g., drones) or battery-powered cells to green edge clouds. We perform an in-depth experimental analysis of the energy consumption of virtualized Base Stations (vBSs) and render two conclusions: (i) characterizing performance and power consumption is intricate as it depends on human behavior such as network load or user mobility; and (ii) there are many control policies and some of them have non-linear and monotonic relations with power and throughput. Driven by our experimental insights, we argue that machine learning holds the key for vBS control. We formulate two problems and two algorithms: (i) BP-vRAN, which uses Bayesian online learning to balance performance and energy consumption, and (ii) SBP-vRAN, which augments our Bayesian optimization approach with safe controls that maximize performance while respecting hard power constraints. We show that our approaches are data-efficient and have provably performance, which is paramount for carrier-grade vRANs. We demonstrate the convergence and flexibility of our approach and assess its performance using an experimental prototype.

We study physical-layer (PHY) baseband functional split policies in 5G Centralized Radio-Access-Network (C-RAN) architectures that include a central location, the baseband unit (BBU) with some BBU servers, and a set of Base Stations (BSs), the remote radio heads (RRHs), each with a RRH server. Each RRH is connected to the BBU location through a fronthaul link. We consider a scenario with many frame streams at the BBU location, where each stream needs to be processed by a BBU server before being sent to a remote radio-head (RRH). For each stream, a functional split needs to be selected, which provides a way of partitioning the computational load of the baseband processing chain for stream frames between the BBU and RRH servers. For streams that are served by the same BBU server, a scheduling policy is also needed. We formulate and solve the joint resource allocation problem of functional split selection, BBU server allocation and server scheduling, with the goal to minimize total average end-to-end delay or to minimize maximum average delay over RRH streams. The total average end-to-end delay is the sum of (i) scheduling (queueing) and processing delay at the BBU servers, (ii) data transport delay at the fronthaul link, and (iii) processing delay at the RRH server. Numerical results show the resulting delay improvements, if we incorporate functional split selection in resource allocation.

This paper presents RapidRider, the first WiFi backscatter system that takes uncontrolled OFDM WiFi signals, e.g., 802.11a/g/n, as excitations and efficiently embeds tag data at the single-symbol rate. Such design brings us closer to the dream of pervasive backscatter communication since uncontrolled WiFi signals are everywhere. Specifically, we show that RapidRider can demodulate tag data for each OFDM symbol while previous systems rely on multi-symbol demodulation. Further, we design deinterleaving-twins decoding that enables RapidRider to use any uncontrolled WiFi signals as carriers. We prototype RapidRider using FPGAs, commodity radios, and USRPs. Comprehensive evaluations show that RapidRider's maximum throughput is 3.92x and 1.97x better than FreeRider and MOXcatter. To accommodate cases where there is only one receiver available, we design RapidRider+ that can take productive data and tag data on the same packet. Results demonstrate that it can achieve an aggregated goodput of productive and tag data around 1 Mbps on average.

We introduce a novel representation of packet classifiers allowing to operate on partially available input data varying dynamically. For a given packet classifier, availability of fields or complexity of field computations, and free target specific resources, the proposed infrastructure computes a classifier representation satisfying performance and robustness requirements. We show the feasibility to reconstruct a classification result in this noisy environment, allowing for the improvement of performance and the achievement of additional robustness levels of network infrastructure. Our results are supported by extensive evaluations in various settings where only a partial input is available.