Using programmable network devices to aid in-network machine learning has been the focus of significant research. However, most of the research was of a limited scope, providing a proof of concept or describing a closed-source algorithm. To date, no general solution has been provided for mapping machine learning algorithms to programmable network devices. In this paper, we present Planter, an opensource, modular framework for mapping trained machine learning models to programmable devices. Planter supports a wide range of machine learning models, multiple targets and can be easily extended. The evaluation of Planter compares different mapping approaches, and demonstrates the feasibility, performance, and resource efficiency for applications such as anomaly detection, financial transactions, and quality of experience. The results show that Planterbased in-network machine learning algorithms can run at line rate, have a negligible effect on latency, coexist with standard switching functionality, and have no or minor accuracy trade-offs.
Since Bluetooth5 standard released in 2016, its usage in commercial electronic products has been increased rapidly and substantially. Comparing to BLE 4.2, Bluetooth5 supports three PHY modes, respectively 2M, 1M and Coded PHY mode, providing a higher throughput and a wider range. Whereas there is a trade-off between its throughput and coverage. When the connection is established, the PHY mode is commonly preconfigured and fixed. This rigid design limits the flexibility in offering dynamic throughput and coverage. Therefore, we propose a method termed AptBLE, that switches the PHY mode in Bluetooth5 adaptively by considering the Received Signal Strength Indicator (RSSI) level. Specifically, we optimise the RSSI threshold for different PHY modes using the K-means clustering algorithm. Moreover, based on AptBLE, we further enable the Data Length Extension (DLE) feature and term the improved method as AptBLEM. We implement AptBLE (M) on the boards and test in indoor environment. The experimental results show that, AptBLE is more flexible, robust and outperforms the original fixed PHY mode in terms of throughput and transmission range. Furthermore, AptBLEM can triple the throughput than AptBLE, with a maximum throughput value in 1035Kbps and 42m range in indoor environment.
Internet traffic is predicted to increase fast over the next years. A large portion of it will be generated by mobile video services. Such a data explosion puts higher requirements on the capacity of the mobile network. Deploying more bandwidth resources to increase the network capacity is one solution, but it also means high cost. Mobile Edge Caching (MEC) is a new solution put forward these years to deal with the drastic growth of video data by bringing the video resources close to users at the edge cache. Researches have been done on the design of MEC, and implementing it in an emulator is one of the ways to verify the design. An emulator can provide real-case protocol implementation and more credible results compared with simulator. This paper studies the LTE emulators available in the market and proposes an MEC testbed based on the OpenAirInterface platform. Two use cases of the testbed are demonstrated and their performances are evaluated separately.
Index Terms-Testbed, mobile edge caching, LTE emulation, open air interface.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.