A Computing Resource Management Framework for Software-Defined Radios Vuk Marojevic, Student Member, IEEE, Xavier Reve ´s Balleste ´, Member, IEEE, and Antoni Gelonch Abstract—Software-defined radio (SDR) is an emerging concept that leverages the design of software-defined and hardware- independent signal processing chains for radio communications. It introduces flexibility to wireless systems, facilitating the dynamic switch from one radio access technology to another or, in other words, the de and reallocation of computing resources from one SDR application to another. This paper introduces an SDR computing resource management framework. It accounts for several SDR system characteristics, including real-time computing requirements, limited computing resources, and heterogeneous multiprocessor platforms. The framework features the t w -mapping, a dynamic mapping algorithm that is apt for many cost functions and radio scenarios. The cost function proposal dynamically manages the available computing resources to satisfy the SDR computing constraints. Two SDR scenarios, based on representative SDR platforms and processing chains, and the corresponding simulation results demonstrate the framework’s relevance and suitability for SDRs. Index Terms—Computing resource management, framework, heterogeneous computing, mapping, scheduling, software-defined radio (SDR), system modeling. Ç 1 INTRODUCTION T HE software radio concept was introduced in the mid-1990s and characterizes those transmitters and receivers (transceivers) that implement the entire signal processing chain in software [1], [2]. Software-defined radio (SDR) can be considered a generalization of software radio because it characterizes a transceiver that implements one or more signal processing blocks in software [3]. SDR introduces flexibility to wireless systems: It permits the adjustment or switching of a terminal’s radio access technology (RAT) implementation to adapt to changes in the radio environ- ment of today and tomorrow. For about a decade, SDR-related research along the whole line between the mobile terminal (MT) transceiver and the core network has been ongoing [4], [5], [6], [7]. It is motivated by the evolution of information technology: The introduction of new RATs, such as the universal mobile telecommunications system (UMTS) or the IEEE 802.11 family of wireless local area networks, the required compatibility with the existing ones, including global system for mobile communications (GSM) and general packet radio service (GPRS), and the increasing demand for new and differentiated user services call for flexible transceiver solutions. For the above reasons, the flexibility of general-purpose processors (GPPs), digital signal processors (DSPs), field-programmable gate arrays (FPGAs), picoArrays [8], networks-on-chip (NoCs) [9], or multiprocessor systems-on- chip (MP-SoCs) [10] is gaining interest over the energy efficiency of application-specific integrated circuits (ASICs) [11], [12], [13]. State-of-the-art reconfigurable devices, including arrays of processors, offer high computing capacities at moderate power consumptions. This permits the extension of the digital and reconfigurable radio part while reducing the analog and nonreconfigurable circuits. An SDR processing chain (SDR application or waveform) is the part of an SDR transceiver that is implemented in software. It may be understood as a set of concurrent processes that continuously process and propagate real- time data. Such a processing chain is not specifically tailored, but, rather, executable on any general-purpose platform with sufficient computing capacity. Because of these similarities between future SDR applications and platforms and today’s general-purpose computing applica- tions and platforms, we consider general-purpose comput- ing methods practical for SDR systems. We particularly believe that the introduction of appropriate mapping and scheduling techniques, which are essential for the dynamic switch between RATs, will leverage the design of SDR platforms and applications. Mapping describes the process of assigning software modules to hardware resources, whereas scheduling determines the execution times of these modules. We consider them as two complementary methods for computing resource management. Wireless or SDR systems, however, reveal specific aspects, essentially regarding flexibility and efficiency, which have not been jointly considered so far in hetero- geneous computing. These are given as follows: 1. time slot based division of the transmission medium (radio time slot), 2. continuous data transmission and reception, 3. RAT-specific quality-of-service (QoS) targets, IEEE TRANSACTIONS ON COMPUTERS, VOL. 57, NO. 10, OCTOBER 2008 1399 . V. Marojevic and A. Gelonch are with the Departament de Teoria del Senyal i Comunicacions (TSC), Universitat Polite`cnica de Catalunya, Av/ Canal Olı´mpic s/n, 08860 Castelldefels, Barcelona, Spain. E-mail: {marojevic, antoni}@tsc.upc.edu. . X.R. Balleste´is with Gigle Semiconductor S.I., Barcelona, Spain. E-mail: xavier.reves@tsc.upc.edu. Manuscript received 6 July 2007; revised 3 Jan. 2008; accepted 13 Mar. 2008; published online 19 May 2008. Recommended for acceptance by V. Barbosa. For information on obtaining reprints of this article, please send e-mail to: tc@computer.org, and reference IEEECS Log Number TC-2007-07-0308. Digital Object Identifier no. 10.1109/TC.2008.83. 0018-9340/08/$25.00 ß 2008 IEEE Published by the IEEE Computer Society