We present a novel data-driven approach to perform smooth Wi-Fi/cellular handovers on smartphones. Our approach relies on data provided by multiple smartphone sensors (e.g., Wi-Fi RSSI, acceleration, compass, step counter, air pressure) to predict Wi-Fi connection loss and uses Multipath TCP to dynamically switch between different connectivity modes. We train a random forest classifier and an artificial neural network on real-world sensor data collected by five smartphone users over a period of three months. The trained models are executed on smartphones to reliably predict Wi-Fi connection loss 15 seconds ahead of time, with a precision of up to 0.97 and a recall of up to 0.98. Furthermore, we present results for four DASH video streaming experiments that run on a Nexus 5 smartphone using available Wi-Fi/cellular networks. The neural network predictions for Wi-Fi connection loss are used to establish MPTCP subflows on the cellular link. The experiments show that our approach provides seamless wireless connectivity, improves quality of experience of DASH video streaming, and requires less cellular data compared to handover approaches without Wi-Fi connection loss predictions.

This paper presents MACI, the first bespoke framework for the management, the scalable execution, and the interactive analysis of a large number of network experiments. Driven by the desire to avoid repetitive implementation of just a few scripts for the execution and analysis of experiments, MACI emerged as a generic framework for network experiments that significantly increases efficiency and ensures reproducibility. MACI incorporates and integrates established simulators and analysis tools to foster rapid but systematic network experiments.

Multipath TCP (MPTCP) is a recent TCP evolution that uses multiple TCP subflows and thereby different network paths for a single MPTCP connection. The MPTCP scheduler has a significant impact on the overall performance of MPTCP, as it maps outgoing packets on TCP subflows. In this paper, we identify limitations of today's MPTCP schedulers for thin streams. We present a novel MPTCP scheduler which overcomes today's limitations for thin streams by using i) active probing of unused subflows and ii) timely one-way delay information. A systematic evaluation within our MPTCP Linux Kernel implementation shows that our novel scheduler outperforms established MPTCP schedulers with regard to application-layer round-trip time without sacrificing efficiency and throughput.

QUIC is the emerging transport layer protocol, providing encrypted, stream-multiplexed, low-latency data transfer. In this paper, we propose multipath-enabled QUIC (MPQUIC) to leverage multiple network interfaces, such as WiFi and LTE on today's mobile devices. We show how our MPQUIC design conceptually evolves beyond existing multipathing protocols, such as MPTCP, as it provides fine-grained stream-to-path scheduling, reduced head-of-line blocking, and faster subflow establishment. We present an userland implementation of MPQUIC that is deployable without operating system changes. Our evaluation results show that MPQUIC increases throughput in comparison to traditional QUIC, TCP and even the currently de facto multipath transport protocol MPTCP. First real world measurements confirm that MPQUIC is deployable in the Internet to reduce download times. Moreover, we show that MPQUIC's conceptual advantages over MPTCP efficiently reduce head-of-line blocking in heterogeneous environments. With multipathing support, QUIC is ready to become the universal stream transport protocol in today's Internet.

This paper presents a crowdsourced dataset of a large-scale event with more than 1000 measuring participants. The detailed dataset consists of various location data and network measurements of all national carrier collected during a four-day event. The concentrated samples for this short time period enable detailed analysis, e.g., by correlating movement patterns and experienced network conditions.

Fork-Join (FJ) queueing models capture the dynamics of system parallelization under synchronization constraints, for example, for applications such as MapReduce, multipath transmission and RAID systems. Arriving jobs are first split into tasks and mapped to servers for execution, such that a job can only leave the system when all of its tasks are executed. In this paper, we provide computable stochastic bounds for the waiting and response time distributions for heterogeneous FJ systems under general parallelization benefit. Our main contribution is a generalized mathematical framework for probabilistic server scheduling strategies that are essentially characterized by a probability distribution over the number of utilized servers, and the optimization thereof. We highlight the trade-off between the scaling benefit due to parallelization and the FJ inherent synchronization penalty. Further, we provide optimal scheduling strategies for arbitrary scaling regimes that map to different levels of parallelization benefit. One notable insight obtained from our results is that different applications with varying parallelization benefits result in different optimal strategies. Finally, we complement our analytical results by applying them to various applications showing the optimality of the proposed scheduling strategies.

Multipath TCP enables remarkable optimizations for throughput, load balancing, and mobility in today's networks. The design space of Multipath TCP scheduling, i.e., the application-aware mapping of packets to paths, is largely unexplored due to its inherent complexity. Evidence in this paper suggests that an application-aware scheduling decision, if leveraged right, pushes Multipath TCP beyond throughput optimization and thereby provides benefits for a wide range of applications.

The current body of research on Dynamic Adaptive Streaming over HTTP (DASH) contributes various adaptation algorithms aiming to optimize performance metrics such as the Quality of Experience. Intuitively, the heterogeneity of the streaming environment and the underlying technologies lead many of the developed approaches to possess clear performance affinities denoted here as sweet spots. We observe, however, that systematic comparisons of these algorithms are usually conducted within homogeneous player environments.

More and more Internet-enabled devices, such as server instances or smartphones, have multiple network interfaces. Multipath TCP (MPTCP) has proven to increase bandwidth for these devices, while remaining compatible with the existing network infrastructure and applications. For interactive applications and services, however, low latency and low jitter often is more important than bandwidth. In this paper, we rethink the MPTCP approach, focusing on end-to-end latency and jitter. We propose ReMP TCP, an MPTCP extension that sends data redundantly over multiple paths in the network. Exchanging bandwidth for latency, this approach guarantees the lowest possible latency in existing best-effort networks. The integration into the MPTCP protocol provides benefits such as transparent end-to-end connection establishment, multipath-enabled congestion control, and the prevention of head of line blocking. We discuss end-to-end latency in multipath environments considering both queuing delays and packet drops. Further, we evaluate the performance of our ReMP TCP Linux Kernel implementation for data center and mobile scenarios in Mininet and real world experiments. We show for a real world mobile scenario in a stressed environment that ReMP TCP can halve the average round-trip time and reduce its standard deviation by a factor of 19.

Network emulations are widely used in the networking community. The network emulator Mininet recently gained popularity, as it allows running real Linux applications on top of an emulated network. The specification of the network includes the topology as well as static bandwidth, latency, and packet drops probability parameters. Even though evaluations with static parameters provide useful insights, real world measurements show dynamically changing bandwidths, posing special challenges that need to be addressed in network research.

Due to complex interdependencies and feedback loops between network layers and nodes, the development of adaptive applications is difficult. As networking applications respond nonlinearly to changes in the environment and adaptations, defining concrete adaptation rules is nontrivial. In this paper, we present the offline learner Fossa, which uses genetic programming to automatically learn suitable Event Condition Action (ECA) rules. Based on utility functions defined by the developer, the genetic programming learner generates a multitude of rule sets and evaluates them using simulations to obtain their utility. We show, for a concrete example scenario, how the genetic programming learner benefits from the clear model of the ECA rules, and that the methodology efficiently generates ECA rules which outperform nonadaptive and manually tuned solutions.

The handover from WiFi to mobile networks is known to lead to TCP connection drops due to changing IP addresses. Multipath TCP (MPTCP), a recent TCP extension, enables a transparent mobile handover by combining subflows on multiple interfaces, such as WiFi and LTE, to one logical connection. MPTCP provides multiple handover modes, which differ in their energy consumption and the performance during the handover. The Full-MPTCP mode uses permanently both WiFi and the mobile network, which increases energy consumption. The Single-Path mode establishes the mobile network connection after the WiFi connection broke, which leads to a short performance degradation. In this paper, we argue that this trade-off is not necessary. We propose to use the available (sensor) information to forecast the mobile handover. This allows switching to the Full-MPTCP mode before the WiFi connection breaks, providing both low energy consumption and high performance during the handover. For a first experimental evaluation, we use a declining WiFi link quality to forecast a handover. Our real world measurements show that both low energy consumption and high performance during the handover are possible at the same time.