Progressive Multicore RLNC Decoding With Online DAG Scheduling

Abstract

A complete generation of packets coded with Random Linear Network Coding (RLNC) can be quickly decoded on a multicore system by scheduling the involved matrix block operations in parallel with an offline (pre-recorded) directed acyclic graph (DAG). The waiting for a complete generation of packets can be avoided with progressive RLNC decoding that commences the decoding (and can decode some packets) before all packets in a generation have been received. This article develops and evaluates a novel progressive RLNC decoding strategy based on the principle of DAG scheduling of parallel matrix block operations. The novel strategy involves helper matrices for conducting the Gauss Jordan elimination based on rows of blocks of matrix elements. The matrix block computations are dynamically scheduled by an online DAG which permits branching, e.g., to skip unnecessary matrix block operations. The throughput and delay of the novel progressive RLNC decoding strategy are evaluated with experiments on two heterogeneous multicore processor boards. The novel progressive RLNC decoding achieves throughput levels on par with stateof-the-art non-progressive (full-generation) RLNC decoding and achieves three times higher throughput than the fastest (highest-throughput) known progressive RLNC decoder for small generation sizes and short data packets. Also, our progressive RLNC decoding greatly reduces receiver delays for moderate to large generation sizes; the delay reductions are particularly pronounced when a low-delay RLNC version is employed (e.g., reduction to one tenth of the non-progressive decoding delay for a generation size of 256 packets). INDEX TERMS Directed acyclic graph (DAG), helper matrix, heterogeneous multicore architecture, online scheduling, parallel computing, random linear network coding (RLNC).

DSEP Fulcrum: Dynamic Sparsity and Expansion Packets for Fulcrum Network Coding

“…Furthermore, the DSEP Fulcrum coding in this article has been limited to non-systematic coding (specifically, only the outer coding has been systematic, but the inner coding has been non-systematic); an important future research direction is to extend DSEP Fulcrum coding to fully systematic coding [15], [33]- [35] so as to allow the immediate transmission of the uncoded source packets without any encoding delay. Additionally, the speed-up of the DSEP Fulcrum encoding and decoding computations through specialized parallel computing strategies on multicore plat-forms [16]- [19] should be explored in future research. Moreover, the sparsity and extra coded packets could be adapted according to the loss conditions, e.g., high-loss periods could delay the increase of the receiver rank.…”

“…networks [3], cellular and radio access networks [4], [5], vehicular and wireless sensor networks [6], [7], and general wireless networks [8]- [12], as well as unreliable complex information technology infrastructures, such as data caching infrastructures [13]- [15]. One main hurdle that prevents the widespread adoption of RLNC in communication networks and information technology systems is the computationally highly demanding matrix multiplication and matrix inversion required for RLNC decoding [16]- [19]. While the computational capabilities of network nodes are generally increasing and some senders and receivers have abundant computational capabilities [20], [21], a wide range of senders and receivers will continue to have limited computational capabilities for the foreseeable future.…”

“…We have initially considered an end-to-end (one-hop) coding scenario, where the source node (sender) encodes and the destination node (receiver) decodes, without recoding at intermediate network nodes. The decoding employed a progressive decoder (on-the-fly version of Gauss Jordan algorithm [97]) that starts the decoding (coding coefficient elimination) with the first received packet [16], [19]. The encoding throughput is defined as the amount of payload data in a generation, i.e., n× data packet size, divided by the encoding computation time for n + r coded packets, including the outer encoding and the inner encoding.…”

Hardware-Accelerated Platforms and Infrastructures for Network Functions: A Survey of Enabling Technologies and Research Studies

“…The IS extensions have been utilized to accelerate hashing NFs [109], [110] and dynamic task scheduling [111]. Similar IS extensions have accelerated the complex network coding function [112], [113], [193] in a hardware design [114].…”

Device-Enhanced MEC: Multi-Access Edge Computing (MEC) Aided by End Device Computation and Caching: A Survey

“…Wireless communication is generally error prone and requires safeguards at the physical, e.g., forward error correction coding, and link layers, e.g., automatic repeat request retransmissions, to ensure reliability. An emerging coding technique that appears well suited to be explored for the highly heterogeneous device-enhanced MEC in future research is network coding [275]- [280]. Network coding eliminates the coordination that is required between sender and receiver in many conventional coding techniques.…”