Edge computing aims to enable applications with stringent latency requirements, e.g., augmented reality, and tame the overwhelming data streams generated by IoT devices. A core principle of this paradigm is to bring the computation from a distant cloud closer to service consumers and data producers. Consequentially, the issue of edge computing facilities' placement arises. We present a comprehensive analysis suggesting where to place general-purpose edge computing resources on an Internet-wide scale. We base our conclusions on extensive real-world network measurements. We perform extensive traceroute measurements from RIPE Atlas to datacenters in the US, resulting in a graph of 11K routers. We identify the affiliations of the routers to determine the network providers that can act as edge providers. We devise several edge placement strategies and show that they can improve cloud access latency by up to 30%.

In the early days of cloud computing, datacenters were sparsely deployed at distant locations far from end-users with high end-to-end communication latency. However, today’s cloud datacenters have become more geographically spread, the bandwidth of the networks keeps increasing, pushing the end-users latency down. In this paper, we provide a comprehensive cloud reachability study as we perform extensive global client-to-cloud latency measurements towards 189 datacenters from all major cloud providers. We leverage the well-known measurement platform RIPE Atlas, involving up to 8500 probes deployed in heterogeneous environments, e.g., home and offices. Our goal is to evaluate the suitability of modern cloud environments for various current and predicted applications. We achieve this by comparing our latency measurements against known human perception thresholds and are able to draw inferences on the suitability of current clouds for novel applications, such as augmented reality. Our results indicate that the current cloud coverage can easily support several latency-critical applications, like cloud gaming, for the majority of the world’s population.

We consider large scale Internet of Things applications requesting data from physical devices. We study the problem of timely dissemination of sensor data towards applications with freshness requirements by means of a cache. We aim to minimize direct access to the possibly battery powered physical devices, yet improving Age of Information as a data freshness metric. We propose an Age of Information-aware scheduling policy for the physical device to push sensor updates to the caches located in cloud data centers. Such policy groups application requests based on freshness thresholds, thereby reduces the number of requests and threshold misses, and accounts for delay variation. The policy is incrementally introduced as we study its behavior over ideal and more realistic communication links with delay variation. We numerically evaluate the proposed policy against a simple yet widely used periodic schedule. We show that our informed schedule outperforms the periodic schedule even under high delay variations.

Due to their popularity, large coverage areas, and diversity of radio technologies, Low Power Wide Area Networks (LPWAN) will experience high levels of cross-technology interference. In this work, we investigate the impact of LoRa interference on IEEE 802.15.4 (Wi-SUN) networks. Using hardware experiments, we characterize the interaction between these two very different radio technologies. In particular, we observe that the IEEE 802.15.4 Clear Channel Assessment (CCA) mechanism does not reliably detect interfering LoRa transmissions. We therefore propose an enhanced CCA mechanism based on a Multi-Layer Perceptron classifier and show that it significantly reduces the number of unsuccessful transmissions, while remaining compatible with the IEEE 802.15.4 standard.

The large number and wide diversity of IoT networks operating in unlicensed spectrum will create a complex and challenging interference environment. To avoid a 'tragedy of the commons', networks may need to more explicitly coordinate their use of the shared channel.

We present some preliminary results on LoRaWAN and IEEE 802.15.4-SUN interference in urban environments. The results are based on a simple simulation that is parameterized using PHY layer measurements of controlled interference scenarios.

We describe a testbed for studying battery discharge behavior and the lifetime of wireless devices under controlled temperature conditions and present preliminary measurement results.

The rapid growth of new radio technologies for Smart City/Building/Home applications means that models of cross-technology interference are needed to inform the development of higher layer protocols and applications. We systematically investigate interference interactions between LoRa and IEEE 802.15.4g networks. Our results show that LoRa can obtain high packet reception rates, even in presence of strong IEEE 802.15.4g interference. IEEE 802.15.4g is also shown to have some resilience to LoRa interference. Both effects are highly dependent on the LoRa radio's spreading factor and bandwidth configuration, as well as on the channelization. The results are shown to arise from the interaction between the two radios' modulation schemes. The data have implications for the design and analysis of protocols for both radio technologies.

Sensor networks that operate in the unlicensed 2.4~GHz frequency band suffer cross-technology radio interference from a variety of devices, e.g., Bluetooth headsets, laptops using WiFi, or microwave ovens. Such interference has been shown to significantly degrade network performance. We present SoNIC, a system that enables resource-limited sensor nodes to detect the type of interference they are exposed to and select an appropriate mitigation strategy. The key insight underlying SoNIC is that different interferers disrupt individual 802.15.4 packets in characteristic ways that can be detected by sensor nodes. In contrast to existing approaches to interference detection, SoNIC does not rely on active spectrum sampling or additional hardware, making it lightweight and energy-efficient.

We have developed a framework for capturing various performance related sensor data from elite-level cross-country skiers and feeding this information to a set of bespoke analysis tools. The core of the system involve tri-axle accelerometers for skier's upper body movement and an analysis tool which allow us to classify which technique they are using, and how well it is being performed. Other sensor data, such as physiological and positioning information, is used to provide context to the classification.

In the emerging Delay Tolerant Networks (DTNs) content dissemination platform, mobile nodes opportunistically exchange content as they meet, with the intent of sharing content among nodes with common interests. During a meeting, nodes can exchange both content of direct interest to themselves as well as content that is of interest to a larger set of nodes that may be encountered in the future. The utility of a DTN is governed by the content exchange opportunity (the amount of content that can be exchanged during a meeting) as well as the selection of content to be exchanged in order to maximise the interest nodes will have in information they are exposed to. Considering that there is a cost associated with the content exchange (e.g. battery usage, buffer occupancy or consumed transmission opportunity) the aim for nodes participating in content dissemination should be to maximise their payoff. In this paper, we contribute a generic framework for describing the characteristics of content exchange among participating nodes in a network. We incorporate a distributed information popularity measurement and the pairwise interaction of nodes modelled as a bargaining problem. The outcome of this process is the fair split up of transmission opportunity as a network resource and the selection of content objects to exchange in order to maximise the nodes' payoff. The framework is generally intended as a capstone for investigation of content dissemination properties and various content exchange strategies in a DTN. The paper further presents experiments conducted to validate the function and correctness of the proposed framework.

Many sensor network application scenarios include mobile nodes, such as a moving sink. Evaluation of such applications in a testbed is challenging as the testbed has to support mobile nodes. Here, we present Sensei-UU, a sensor network testbed that supports mobile sensor nodes. The testbed is inexpensive, relocatable and possible to reproduce by other researchers. Its primary design objectives are to support experiments with repeatable mobility and relocating the testbed deployment to different locations. Mobile sensor nodes are carried by robots that use floor markings for navigation and localization. The testbed can be used to evaluate applications in which sensor nodes move in the order of meters rather than millimeters, e.g. when a human carries a mobile phone that collects data while passing stationary sensor nodes. To investigate the repeatability of robot movements, we measure the achieved precision and the timing of the robots, and find that our robot localization is accurate to ±1 cm. Furthermore, we investigate variations in radio signal strengths between mobile and stationary nodes. We study the impact of imprecise movements, external sources of interference and environmental influences. We conclude that Sensei-UU supports experiments in which these variations are acceptably low to capture small-scale fading phenomena in IEEE 802.15.4.

We study the packet dispersion phenomenon that a traffic flow experiences when it passes through a router. We show that when there are competing flows and the router schedules packets first-come-first-served (FCFS), the dispersion is not described well by the bottleneck spacing effect. We therefore introduce the term FCFS-spacing effect. We also show that for a router implementing weighted fair queuing, dispersion is due either to the bottleneck spacing effect or the FCFS spacing effect. The results from real measurements and simulations corroborate each other. Finally, we discuss the implications of FCFS packet dispersion on TCP's self-clocking mechanism.

Web servers must be protected from overload since server overload leads to low server throughput and increased response times experienced by the clients. Server overload occurs when one or more server resources are overutilized. We present an adaptive architecture that performs admission control based on the expected resource consumption of requests. By dynamically setting the maximum rate of accepted requests, we avoid overutilization of the critical resources. We also provide mechanisms for service differentiation. We present our admission control architecture and experiments that show that it can sustain low response times and high throughput for premium clients even during high load.

We present a network friendly bandwidth measurement method, TOPP, that is based on active probing and includes analysis by segmented regression. This method can estimate two complementing available bandwidth metrics in addition to the link bandwidth of the congested link. Contrary to traditional packet pair estimates of the bottleneck link bandwidth, our estimate is not limited by the rate at which we can inject probe packets into the network. We also show that our method is able to detect bottlenecks that are invisible to methods such as the C-probe. Further more, we describe scenarios where our analysis method is able to calculate bandwidth estimates for several congested hops based on a single end-to-end probe session.

We review previous work on the applicability and performance of integrated layer processing (ILP). ILP has been shown to clearly improve computer communication performance when integrating simple data manipulation functions, but the situation has been less clear for more complex functions and complete systems. We discuss complications when applying ILP to protocol stacks, the requirements of ILP on the communication subsystem, caching aspects, the importance of the processor registers, and a model for predicting the performance of data manipulation functions. We conclude that the main drawback of ILP is its limited applicability to complex data manipulation functions. The performance to expect from an ILP implementation also depends heavily on the protocol architecture and the host system architecture.

Two major performance bottlenecks in multiprocessor execution of protocols are contention for shared memory and for locks. Locks are used to protect shared messages and/or shared protocol state in a memory shared by competing processors. Mutual exclusion by locking can be costly, in terms of both lock contention and memory contention, if the parallel protocol code frequently accesses shared state and data. This paper presents a queueing network model for performance predictions of shared-memory multiprocessor protocol executions. Predictions from this model are compared to performance measurements from a multiprocessor implementation of two commonly used communication protocol stacks, transmission control protocol/Internet protocol (TCP/IP)/Ethernet and user datagram protocol/Internet protocol (UDP/IP)/Ethernet. These stacks are implemented on a parallelized version of the x-kernel protocol environment from the University of Arizona. A "processor-per-message" paradigm is used to partition the load among the processors. The measured speedups for the parallel implementations relative to the sequential ones are more than 11 times for UDP (using 20 processors) and three times for TCP (using five processors) on a sequent symmetry. We show that the model accurately captures the effects of lock and memory contention in our shared-memory multiprocessor and predicts the performance with a discrepancy of less than 10%.

The implementation of protocols, such as TCP/IP, and their integration into the operating system environment is crucial for protocol performance. Putting TCP on high-speed networks, e.g., ATM, with large maximum transmission units causes the TCP maximum segment size to be relatively large. What Nagle's algorithm considers a "small" segment is no longer small, which affects the TCP performance. The authors report on TCP/IP throughput and RPC response time performance measurements for various sizes of send and receive socket buffers, using the Sparc10 architecture machines Axil 311/5.1 running SunOS 4.1.3 connected to a FORE System's ATM network. For some common combinations of socket buffer sizes the authors observe a dramatic performance degradation to less than 1% of expected throughput and to one order-of-magnitude longer response time than expected. The performance degradation is caused by a deadlock situation in the TCP connection which is resolved by the 200 ms spaced timer generated TCP delayed acknowledgment. The authors explain the causes of the deadlock situations, and discuss means to avoid or prevent them.>

We investigate how to exploit shared memory multiprocessors for parallel protocol processing. We present a multiprocessor implementation of the x-kernel protocol environment from the University of Arizona. A processor-per-message paradigm is used to partition the work over processors. Locks are used to protect shared protocol state and data. Mutual exclusion by locking can be costly if the parallel protocol code frequently accesses shared state and data. This paper addresses the effect of locking on performance.The x-kernel is implemented in user address space on a Sequent Symmetry. Performance results of UDP/IP and TCP/IP are given, and we compare them to performance predictions from simulations. The measured speedups for the parallel implementations relative to the sequential ones are more than 12 times for UDP and 3 times for TCP. These limits are set by locking and machine contention.

The authors present SPIMS, a tool which provides an environment where it is easy to develop, specify, and execute communication measurements that can be used for 'benchmarking'. The tool measures user-oriented performance parameters, such as throughput and response time, without relying on special hardware or operating system support. The measurements are specified in a protocol-independent specification language, which is presented in some detail. The usefulness of the tool is demonstrated by measurements on the ISODE OSI (FTAM) stack and the Arpa (FTP/TCP) protocols on Sun, MicroVAX, Apollo, IBM PC/RT and DS90 machines running Unix.>