Achuthan Paramanathan

Lean and mean: network coding for commercial devices

With its ability to reduce the number of transmissions in lossy networks as well as its potential to simplify the design and required signaling of communication protocols, network coding has emerged as an attractive solution to harness the power of wireless and cooperative networks in order to provide higher throughput and lower energy expenditure. This article shows that network codings complexity is not an issue for current mobile devices even without hardware acceleration. We provide real-life measurements of energy savings gains of two design styles of network coding, namely, inter- and intra-session network coding using commercial platforms, including Open-Mesh routers and various mobile phones. We demonstrate that the energy per bit invested in coding/decoding operations can be several orders of magnitude smaller than that used for transmission/reception, while also maintaining processing speeds as high as several hundreds of Mb/s or even several Gb/s depending on the device and coding configuration used.

Sharing the Pi: Testbed Description and Performance Evaluation of Network Coding on the Raspberry Pi

This paper presents the design and performance evaluation of an inexpensive testbed for network coding protocols composed of Raspberry Pis. First, we show the performance of random linear network coding primitives on the Raspberry Pi in terms of processing speed and energy consumption under a variety of configuration setups. Our measurements show that processing rates of up to 230 Mbps are possible with the Raspberry Pi. Also, the energy consumption per bit can be as small as 3 nJ/bit, which is several orders of magnitude smaller than the transmission/reception energy use. Surprisingly, overclocking the Raspberry Pi from 700 MHz to 1000 MHz not only produces an increase in processing speed of up to 68 % for large generation sizes, but also provides a reduction of 64 % in the processing energy per bit for most tested scenarios. Then, we show Raspberry Pi as an inexpensive, viable, and flexible platform to deploy large research networking testbeds for the evaluation of network coding protocols. We propose key parameters and representations to evaluate protocol performance in network nodes as well as validating the testbeds statistics using the case of a one-hop broadcast with random linear network coding, which is well understood in theory.

Energy consumption model and measurement results for network coding-enabled IEEE 802.11 meshed wireless networks

This paper presents an energy model and energy measurements for network coding enabled wireless meshed networks based on IEEE 802.11 technology. The energy model and the energy measurement testbed is limited to a simple Alice and Bob scenario. For this toy scenario we compare the energy usages for a system with and without network coding support. While network coding reduces the number of radio transmissions, the operational activity on the devices due to coding will be increased. We derive an analytical model for the energy consumption and compare it to real measurements for which we build a flexible, low cost tool to be able to measure at any given node in a meshed network. We verify the precision of our tool by comparing it to a sophisticated device. Our main results in this paper are the derivation of an analytical energy model, the implementation of a distributed energy measurement testbed conducting a series of measurements, and finally a comparison of the analytical energy model and the data achieved by our testbed.

On the need of novel Medium Access Control schemes for network coding enabled wireless mesh networks

This paper advocates for a new Medium Access Control (MAC) strategy for wireless meshed networks by identifying overload scenarios in order to provide additional channel access priority to the relay. The key behind our MAC protocol is that the relay will adjust its back off window size according to the incoming and outgoing packet ratio. We describe the new protocol as an extension to the CSMA/CA protocol and implement the protocol on our own hardware platform. By means of our own testbed, we investigate two basic network structures, namely, the two-way relay and the cross topology. It is well known that network coding will improve the throughput in such systems, but our novel medium access scheme improves the performance in the cross topology by another 66 % for network coding and 150 % for classical forwarding in theory. These gains translate in a theoretical gain of 33 % of network coding over classical forwarding when both systems implement the improved MAC. However, our measurement results show an even larger gain for network coding, namely, up to 65 % over forwarding, as it copes better with channel losses under high load scenarios.

Leaner and meaner: Network coding in SIMD enabled commercial devices

Although random linear network coding (RLNC) constitutes a highly efficient and distributed approach to enhance communication networks and distributed storage, it requires additional processing to be carried out in the network and in end devices. For mobile devices, this processing translates into energy use that may reduce the battery life of a device. This paper focuses not only on providing a comprehensive measurement study of the energy cost of RLNC in eight different computing platforms, but also explores novel approaches (e.g., tunable sparse network coding) and hardware optimizations for Single Instruction Multiple Data (SIMD) available in the latest generations of Intel and Advanced RISC Machines (ARM) processors. Our measurement results show that the former provides gains of two-to six-fold from the underlying algorithms over RLNC, while the latter provides gains for all schemes from 2× to as high as 20×. Finally, our results show that the latest generation of mobile processors reduce dramatically the energy per bit consumed for carrying out network coding operations compared to previous generations, thus making network coding a viable technology for the upcoming 5G communication systems, even without dedicated hardware.

Energy and data throughput for asymmetric inter-session network coding

In this paper we investigate the impact of asymmetric traffic patterns on the energy consumption and throughput in a wireless multi hop network. Network coding is a novel technique for communication systems and a viable solution for wireless multi hop networks. State of the art research is mainly focusing on ideal scenarios with symmetric traffic patterns that are not realistic in a real life scenario. The main contribution of this paper is the investigation of the asymmetric traffic patterns in terms of throughput and energy consumption, and a validation of these results by real measurements on commercial platforms. The outcome of this paper confirms the analytical expression, and the results shows that even with a large asymmetric data rate there is a gain in terms of energy consumption and throughput when network coding is applied in compare to the case when network coding is not applied.

On the need for buffer optimization strategies for inter-session network coding

During the past decades inter-session network coding for wireless network has been proven to be an effective tool being capable of providing a many-folded gain in terms of throughput. In order to utilize the full potential of inter-session network coding, any coding node needs to operate in an ideal wireless environment that provides equal flow of incoming packets. However, such an ideal wireless environment is seldom seen in a real world scenario. In this work, we address one of the issue that may arise in a non-ideal wireless environment, namely asynchronous packet arrival at the relay leading to an inefficient coding gain. To overcome this, we present a buffer enhancement structure at the coding node that guarantees an equal packet distribution of the codable flows. This structure is implemented on our test platform and using it, we performed a real-life measurement. Our measurement results show that by ensuring an equal packet distribution at the relay yields an improvement in terms of system throughput in compare to an unequally distributed packet buffer structure.

Coding-aware MAC: Providing channel access priority for network coding with reverse direction DCF in IEEE 802.11-based wireless networks

An important challenge for the implementation of network coding in IEEE 802.11-based wireless networks is to give additional priority for channel access to the relay stations responsible for coding. These relay stations are able to provide more information in a single transmission than those that forward single packets, hence improving throughput and energy efficiency. The Distributed Coordination Function (DCF) of the IEEE 802.11 standard is a contention-based Medium Access Control (MAC) protocol that provides an equal distribution of channel access opportunities for all competing stations. However, the relay station represents a congestion point and additional transmission slots should be assigned to it to increase the overall network performance. To address this issue we investigate a coding-aware MAC protocol, called Reverse Direction DCF (RD-DCF), which enables bidirectional communications between the relay station and another station with a single channel access invocation. This simple and backwards compatible mechanism allows the relay station to transmit a coded packet together with the acknowledgement immediately after receiving a data packet. The simulation results show a gain of up to 130% in terms of both throughput and energy efficiency for RD-DCF with network coding when compared to DCF.

Reliable cooperative information storage in wireless sensor networks

Reliable data retrieval from a sensor network can require a temporal distributed data storage of the measured data among all motes. This is the case when a sink of information, or a gateway, is only occasionally present in the network. We consider a network consisting of a set of motes that in order to save energy independently of each other go into offline mode, switching theirs radios off. We propose a cooperative mechanism for distributed data storage that allows complete data retrieval with high probability. Reliable data retrieval is crucial for data-critical applications. Reliability is achieved by introducing redundancy in the system by applying Reed-Solomon scheme for data encoding. This solution can tolerate the situation where a certain number of devices in the sensor cluster is unavailable at the moment of information retrieval. A probabilistic analytical model is developed to investigate the performance of the proposed algorithm. This model is verified by simulation and prototype testing. Prototyping is done on the OpenSensor platform developed at Aalborg University. Performance results indicate that the proposed cooperative data distribution method outperforms the non-cooperative one.

Coping with asymmetric losses in CSMA/CA: Inter-session network coding performance evaluation and adaptive protocol design

Inspired by the discrepancy between past theoretical analysis and real measurements for high-load scenarios for inter-session network coding, we pinpoint and analyze the source of this discrepancy in wireless networks implementing a CSMA/CA medium access scheme. Our analysis shows that CSMA/CA is very sensitive to asymmetric channel losses caused by channel conditions. Leveraging this analysis, we present an adaptive channel priority protocol that copes with asymmetric channel losses while being compatible with CSMA/CA. We implement this protocol and perform real-life measurements that (i) confirm the sensitivity of the CSMA/CA scheme in real implementations, and (ii) shows that our adaptive protocol provides a simple, yet potent mechanism to cope with asymmetric channel losses and ultimately to enhance end-to-end throughput in high-load scenarios.