CoNEXT 2019 Student Workshop Program

Abstract: Recent proliferation of wireless technologies and choices available to user applications have triggered a tremendous wireless demand, and the wireless nodes are expected to dominate the Internet soon. Accommodating this exploding wireless demand with cellular capacity does not seem possible in the long run. As the RF spectrum is getting scarcer and saturated by recent innovations in attaining high spectral efficiency gains, we urgently need innovations that will enable leveraging of new wireless spectrums and substrates in order to respond to the exploding mobile wireless traffic demand. Further, the capacity gap between radio frequency (RF) wireless and optical fiber backbone speeds will remain huge because of the limited availability of RF spectrum. This gap in the “last mile” of the Internet is getting more troublesome as Internet-of-Things (IoT) is becoming a reality and more things around us need wireless connectivity. Enabling optical spectrum in wireless communications is the needed revolution for high-speed mobile networks of the future.

In this talk, I will present our work on exploring the potential for free-space-optics (FSO), a.k.a. optical wireless, in the context of high-speed mobile ad-hoc and opportunistic networking. We introduce basic multi-element building blocks and prototypes for multi-hop FSO-based mobile networking. 3-D spherical structures covered with inexpensive FSO transceivers (e.g., LED and photo-detector pairs) solve issues relevant to mobility and line-of-sight (LOS) management via availability of several transceivers per node. Such structures facilitate electronic LOS tracking (i.e., “electronic steering”) methods instead of traditional mechanical steering techniques used in FSO communications. By abstracting FSO directionality and LOS characteristics, our work also explores issues relating to routing and indoor localization, and develops layer 3 protocols. FSO has been used at high-altitude communications, and this research enables FSO communications at lower-altitudes and in ad-hoc settings with redundancy of cheap optoelectronic components. I will also present our recent efforts on using these multi-element modules for lighting and their potential role in integration of illumination and communication functions.

Abstract: 5G will serve various new use cases that have diverse requirements of multiple resources, e.g., radio, transportation, and computing. Network slicing is a promising technology to slice the network according to the requirements of different use cases. In this work, we present an end-to-end network slicing system that leverages deep reinforcement learning to efficiently orchestrate multiple resources in radio access network, transportation network, and edge computing servers to network slices.

Abstract: In the last decade the ICT community observed a growing popularity of software networking paradigms. This trend consists in moving network applications from static, expensive, hardware equipment (e.g. router, switches, firewalls) towards flexible, cheap pieces of software that are executed on a commodity server. In this context, a server owner may provide the server resources (CPUs, NICs, RAM) for customers, following a Service-Level Agreement (SLA) about clients' requirements. The problem of resource allocation is typically solved by overprovisioning, as the clients' application is opaque to the server owner, and the resource required by clients' applications are often unclear or very difficult to quantify. This paper shows a novel approach that exploits machine learning techniques in order to infer the input traffic load (i.e., the expected network traffic condition) by solely looking at the runtime CPU footprint.

Abstract: Most research papers should have one thing in common: a clear and expressive evaluation of proposed solutions to problems. However, evaluating solutions is interestingly a challenging task: when using human-constructed examples or real-world data, it is difficult to assess to which degree the data represents the input spectrum also of future demands. Moreover, evaluations which fail to show generalization might hide algorithm weak-spots, which could eventually lead to reliability and security issues later on. To solve this problem we propose Toxin, a framework for automated, data-driven benchmarking of, e.g., network algorithms. In a first proof-of-concept implementation, we use Toxin to generate challenging traffic datasets for a data center networking use case.

Abstract: In this paper, we propose CLEO, which is a machine learning approach to equal-cost multipath routing (ECMP) schemes to distribute and balance traffic. ECMP-based traffic load-balancing is widely practiced by datacenters, but hash collision resulting from skewed ECMP hashing makes it difficult to achieve the desired throughputs over paths. Various solutions have been proposed to overcome the performance degradation caused by hash collision, but most of these solutions require modifying packet headers or replacing switches. To solve this problem, CLEO builds a neural-network model that characterizes the ECMP scheme of a switch. The proof-of-concept evaluation shows that CLEO improves the root mean square error fourfold between the desired and real path throughputs.

Abstract: Recently, a Graph Neural Network (GNN) model called RouteNet was proposed as an efficient method to estimate end-to-end network performance metrics such as delay or jitter, given the topology, routing, and traffic of the network. Despite its success in making accurate estimations and generalizing to unseen topologies, the model makes some simplifying assumptions about the network, and does not consider all the particularities of how real networks operate. In this work we extend the architecture of RouteNet to support different features on forwarding devices, specifically we focus on devices with variable queue sizes, and we experimentally evaluate the accuracy of the extended RouteNet architecture.

Abstract: Datacenters in major cloud providers host thousands of competing tenants and applications. Network operators must ensure that available resources are fairly shared and isolated among tenants to meet Service Level Agreements (SLA). Moreover, operators must be able to meet application requirements inside each tenant to provide end-user satisfaction. Providing isolation among tenants, and enforcing application policies require deep, hierarchical policies to isolate tenants and applications separately. Current state of the art approaches cannot enforce deep, hierarchical policies due to the switches' resource limitations. In this paper, we propose Ward, a practical approach to enforce deep hierarchical network policies using packet resubmit in programmable switches. Packet resubmit allows switches to reuse network resources in enforcing complex traffic policies. Our empirical results in a sample hierarchical policy with two levels show that Ward could enforce tenant isolation and strict priority.

Abstract: Several cloud service and data center providers have announced plans to transition towards environmental-friendly green computing. From a computer networking perspective, such transition requires novel routing approaches over traditional networking protocols, more specifically, end-host-driven load balancing. For example, an end-host may issue a request to change its load, based on its current CPU load or energy availability. Our solution, iLoad, collects such requests from end-hosts and computes the routes entirely in the data plane, providing more efficient and reliable operation of a network.

Abstract: Serverless computing is an emerging Cloud paradigm that allows users to claim and pay for resources only when their jobs are executing. While this paradigm offers several advantages, the phenomenon of "cold start" reduces its inherent efficiency with respect to the utilization of compute, storage and network resources that support its existing virtualization deployment systems. We analyze current modes of deployment and identify data similarities across applications. Based on these observations, we propose a new deployment system that is built atop a peer-to-peer network, virtual file-system and content-addressable storage, which will increase compute availability, reduce storage requirement, and prevent network bottlenecks.

Abstract: Web applications utilize in-memory caching systems to reduce the load on backend databases and improve the performance of the system. These cache environments normally host multiple tenants simultaneously, signifying the need to efficiently manage the underlying physical memory allocation in these environments. Off-the-shelf caches statically divide the memory between tenants, which often leads to poor utilization and low hit rates for some of these tenants. In this work, we present Tingo, a multi-tenant cache environment designed to adequately manage the memory allocation of cache tenants to help them adapt to their workloads and optimize their hit ratios, by dynamically reallocating memory pages among them.

Abstract: Today, on-board passengers request Internet-based services, such as video streaming, Voice over IP, etc., at low cost. Many of those services have stringent QoS requirements, e.g., low end-to-end delay. To offer these services efficiently, some problems need to be solved jointly: placement of services (i.e., Virtual Machines (VMs)), at Datacenters (DCs) on the ground, assigning airplanes to services on DCs, and the routing from the flight to the associated VM considering the dynamic position of the flights over time. Further, dynamic VM migrations can be employed for guaranteeing the QoS requirements and/or improving resource utilization. In this work, we introduce and evaluate two heuristic solutions to jointly determine VM placement, routing, and migration decisions for flying airplanes with the objective of minimizing total operational costs. The first results indicate that while reducing the runtime from hours to seconds, the heuristics are able to achieve near-optimal solutions.

Abstract: Emerging communication techniques in terahertz bands, including free-space optical (FSO), have the potential to aid in solving the spectrum scarcity problem by complementing the traditional radio frequency (RF) wireless networks. Line-of-Sight (LOS) link discovery is one of the major limiting factors for these highly directional bands, and this problem of establishing LOS between neighbor nodes becomes more challenging when they have no a-priory knowledge of the location of each other. In this work, we present an asynchronous LOS discovery algorithm using spiral scanning without any synchronization via an RF channel. We also demonstrate the effectiveness of the algorithm through test-bed experiments using a prototype of UAVs equipped with in-band full-duplex (IBFD) FSO transceivers.

Abstract: Low Earth Orbit (LEO) Satellite network have the potential to revolutionize wide-area communications. Due to its location in a complex space environment, LEO satellite network may suffer different failure modes than many traditional networks, and therefore novel approaches are needed to provide them with survivability. In this paper, we propose PKN - a methodology for identifying the key nodes in an LEO Satellite network. We first model the network using time-cumulative graph techniques (C-TVG), which allows us to compute the betweenness centrality of each node in the graph. Finally, we demonstrate the importance of these nodes, showing greater loss in throughout when these nodes are lost.

Abstract: The prediction of individuals' dynamics has attracted significant community attention and has implication for many fields: e.g. epidemic spreading, urban planning, recommendation systems. Current prediction models, however, are unable to capture uncertainties in the mobility behavior of individuals, and consequently, suffer from the inability to predict visits to new places. This is due to the fact that current models are oblivious to the exploration aspect of human behavior. This paper contributes better understanding of this aspect and presents a new strategy for identifying exploration profiles of a population. Our strategy captures spatiotemporal properties of visits -- i.e. a known or new location (spatial) as well as a recurrent and intermittent visit (temporal) -- and classifies individuals as scouters (i.e., extreme explorers), routineers (i.e., extreme returners), or regulars (i.e., with a medium behavior). To the best of our knowledge, this is the first work profiling spatiotemporal exploration of individuals in a simple and easy-to-implement way, with the potential to benefit services relying on mobility prediction.

Abstract: With a rapidly increasing market of millions of devices, the intelligent virtual assistants (IVA) have become a new vector available to exploit security breaches. In this work we approach the third revision of the Amazon Echo ecosystem's device Alexa from a security perspective, focusing our efforts on the interaction between the user and the device. We found the client-server communications to be robust using encryption, but studying the voice message recognition system we discovered a method to execute voice commands remotely, a feature not available by default. This method could be used against the user if an attacker manages to perform a session hijacking attack on the web or mobile clients.

Abstract: This paper proposes transparent blockchain codes to distribute blockchain history. The history data on each node is uncoded (transparent), but entire data obeys the soliton distribution. It not only keeps decentralization, but also brings low bandwidth consumption and good scalability