DR9. 3 Final report of the JRRM and ASM activities

216715 NEWCOM++ DR9.3 216715 NEWCOM++ DR9.3 Final report of the JRRM and ASM activities Contractual Date of Delivery to the CEC: Actual Date of Delivery to the CEC: Editor(s): T0+36 T0+36 Jordi Pérez-Romero (UPC) Participating institutions: CNIT, CNRS, IST-TUL, PUT, UPC. Contributors: António Serrador (IST-TUL), Luisa Caeiro (IST-TUL), Luís M. Correia (IST-TUL), Marco Moretti (CNIT), Emanuel Bezerra (UPC), Pawel Sroka (PUT), Hanna Bogucka (PUT), Miguel López-Benítez (UPC), Anna Umbert (UPC), Ferran Casadevall (UPC), Wassim Jouini (CNRS), Jacques Palicot (CNRS), Christophe Moy (CNRS), Adrian Kliks (PUT), Merouane Debbah (CNRS) Internal Reviewer(s): Sergio Benedetto (ISMB) Workpackage number: WPR9 Nature: R Total Effort Spent: 15 PM Dissemination Level: Public Version: 1 Abstract: This deliverable provides the final report with the summary of the activities carried out in NEWCOM++ WPR9, with a particular focus on those obtained during the last year. They address on the one hand RRM and JRRM strategies in heterogeneous scenarios and, on the other hand, spectrum management and opportunistic spectrum access to achieve an efficient spectrum usage. Main outcomes of the workpackage as well as integration indicators are also summarised. Keyword list: Radio Resource Management (RRM), Joint Radio Resource Management (JRRM), Advanced Spectrum Management (ASM), Game Theory, OFDMA, Cognitive Radio Networks, Opportunistic Spectrum Access. Reference DR9.3 1 / 126 216715 NEWCOM++ DR9.3 Reference DR9.3 2 / 126 216715 NEWCOM++ DR9.3 TABLE OF CONTENTS LIST OF FIGURES ................................................................................................................. 5 LIST OF TABLES ................................................................................................................... 7 LIST OF ACRONYMS............................................................................................................ 8 1 2 INTRODUCTION........................................................................................................... 11 1.1 OBJECTIVES .......................................................................................................................... 11 1.2 DOCUMENT STRUCTURE ...................................................................................................... 11 RRM AND JRRM ALGORITHMS.............................................................................. 13 2.1 INTRODUCTION..................................................................................................................... 13 2.2 JRRM STRATEGIES AND ALGORITHMS ............................................................................. 14 2.2.1 HETEROGENEOUS NETWORKS VHO MODEL ................................................................ 14 2.2.2 JRRM THEORETICAL PARAMETERS’ VARIATION ......................................................... 14 2.3 COOPERATIVE RADIO RESOURCE MANAGEMENT IN MULTI VIRTUAL NETWORKS ...... 16 2.3.1 SIMULATION SCENARIOS ............................................................................................... 20 2.3.2 SIMULATION RESULTS ................................................................................................... 21 2.4 ENERGY EFFICIENCY AWARENESS ON RADIO RESOURCE MANAGEMENT ..................... 25 2.4.1 2.5 SPECTRUM ALLOCATION FOR OFDMA NETWORKS ........................................................ 27 2.5.1 RADIO RESOURCE ALLOCATION IN A SINGLE-CELLULAR SCENARIO ............................. 27 2.5.2 RADIO RESOURCE ALLOCATION IN A MULTI-CELLULAR SCENARIO .............................. 38 2.6 GAME THEORY FOR OPTIMIZATION OF RRM................................................................... 53 2.6.1 2.7 3 METRICS AND COST FUNCTION TO RRM ...................................................................... 25 CRYSTALLIZED RATE REGIONS FOR MIMO TRANSMISSION .......................................... 53 CONCLUSIONS ...................................................................................................................... 73 ADVANCED SPECTRUM MANAGEMENT ............................................................. 76 3.1 INTRODUCTION..................................................................................................................... 76 3.2 MEASUREMENTS OF SPECTRUM AVAILABILITY ................................................................ 76 3.2.1 INTRODUCTION .............................................................................................................. 76 3.2.2 MODELLING SPATIAL SPECTRUM OCCUPANCY .............................................................. 77 3.2.3 SPECTRUM SENSING STUDIES ........................................................................................ 89 3.3 DYNAMIC SPECTRUM ACCESS IN COGNITIVE RADIO NETWORKS ................................... 98 3.3.1 JOINT LEARNING DETECTION FRAMEWORK ................................................................... 98 Reference DR9.3 3 / 126 216715 NEWCOM++ DR9.3 OPPORTUNISTIC SPECTRUM ACCESS WITH SENSING ERRORS: EVALUATION OF UPPER 3.3.2 CONFIDENCE BOUND ALGORITHMS PERFORMANCES .................................................................. 105 3.4 4 CONCLUSIONS .................................................................................................................... 114 CONCLUSIONS ........................................................................................................... 116 4.1 OVERVIEW OF MOST SIGNIFICANT OUTCOMES ................................................................ 116 4.2 INTEGRATION ACTIVITIES ................................................................................................. 117 REFERENCES ..................................................................................................................... 120 Reference DR9.3 4 / 126 216715 NEWCOM++ DR9.3 LIST OF FIGURES Figure 2.1 – Different RAT clusters (reference scenario). .................................................................... 14 Figure 2.2 – JRRM policies impact on LBN blocking.......................................................................... 16 Figure 2.3 – JRRM policies impact on LBN delay. .............................................................................. 16 Figure 2.4 – Inter-VNet RRM and Intra-VNet RRM. ........................................................................... 17 Figure 2.5 – Physical cluster. ................................................................................................................ 20 Figure 2.6 – Simulation scenarios for different VNets composition..................................................... 21 Figure 2.7 – Out of Contract for scenarios 1 and 2. .............................................................................. 22 Figure 2.8 – Service Level for scenarios 1 and 2. ................................................................................. 22 Figure 2.9 – Service Level for scenarios 1 and 3. ................................................................................. 23 Figure 2.10 – Service Level for scenarios 2 and 4. .............................................................................. 23 Figure 2.11 – Maximum Rate capacity without compensation. ............................................................ 24 Figure 2.12 – VNet data rate for scenario 4 with VRRC. ..................................................................... 24 Figure 2.13 – Number of adaptations per type. ..................................................................................... 25 Figure 2.14 – Network, users’ power and information parameters map. .............................................. 26 Figure 2.15 – Impact on fairness of the sub-carrier re-assignment and power re-allocation of the FSRM-P policy.............................................................................................................................. 30 Figure 2.16 – Algorithm 1: Sub-carrier assignment of FSRM-P policy. .............................................. 31 Figure 2.17 – Algorithm 2: Power allocation of FSRM-P policy. ........................................................ 32 Figure 2.18 – Cell fairness index as function of the number of users. .................................................. 35 Figure 2.19 – Total cell throughput as function of the number of users. .............................................. 35 Figure 2.20 – User satisfaction as function of the number of users. ..................................................... 36 Figure 2.21 –Measured spectral efficiency ηm vs. duration of allocation phase Ns ............................. 43 Figure 2.22 –Mean power per cell Pm vs. duration of allocation phase Ns .......................................... 43 Figure 2.23 – Measured spectral efficiency ηm vs. target spectral efficiency η (resource allocation and load control algorithm,no scheduler, same rate for all users)........................................................ 45 Figure 2.24 – Mean power per cell Pm vs. measured spectral efficiency ηm (resource allocation and load control algorithm, no scheduler, same rate for all users)....................................................... 45 Figure 2.25 – Considered cell layout and observation area................................................................... 49 Figure 2.26 – MIMO interference channel: general 2-cell 2-user model (a) and the details representation of the considered 2 2 case (b) ........................................................................... 54 Figure 2.27 – Achievable rate region for the MIMO interference channel - averaged over 2000 channel realizations .................................................................................................................................... 57 Figure 2.28 – Achievable rate region for the MIMO interference channel - one particular channel realization (user two observes strong interference)....................................................................... 58 Figure 2.29 – Achievable rate region for the MIMO interference channel - the transmit power of the first user is twice higher than the transmit power of the second user............................................ 59 Figure 2.30 – Achievable rate region for the precoded MIMO interference channel ........................... 60 Figure 2.31 – Achievable rate region for the OFDM interference channel - results averaged over 2000 channel realizations ....................................................................................................................... 60 Figure 2.32 – Achievable rate region for the OFDM interference channel - one particular channel realization, maximum transmit power of the first user is two times higher than the maximum transmit power of the second user ................................................................................................. 61 Figure 2.33 – Regret-matching learning algorithm ............................................................................... 67 Figure 2.34 – Crystallized rate regions in the interference limited case with marked learned point .... 68 Figure 2.35 – Crystallized rate regions in the noise limited case with marked learned point ............... 68 Figure 2.36 – The convergence of the rate-matching algorithms - user 1............................................. 69 Figure 2.37 – The convergence of the rate-matching algorithms - user 2............................................. 69 Figure 2.38 – SVD-MIMO rate region - channel case 1 ....................................................................... 70 Figure 2.39 – SVD-MIMO rate region - channel case 2 ....................................................................... 70 Figure 2.40 – Achieved average rate versus codebook size .................................................................. 73 Figure 3.1 – Aerial view of UPC’s Campus Nord in urban Barcelona, Spain. ..................................... 79 Reference DR9.3 5 / 126 216715 NEWCOM++ DR9.3 Figure 3.2 – Set of additional measurement locations within UPC’s Campus Nord in urban Barcelona, Spain.............................................................................................................................................. 79 Figure 3.3 – Cross-correlation vs. distance for DCS............................................................................. 81 Figure 3.4 – Covariance vs. distance for DCS. ..................................................................................... 81 Figure 3.5 – Normalised covariance vs. distance for DCS.................................................................... 82 Figure 3.6 – Semivariance vs. distance for DCS................................................................................... 82 Figure 3.7 – Cross-correlation vs. distance for UMTS. ........................................................................ 83 Figure 3.8 – Covariance vs. distance for UMTS. .................................................................................. 83 Figure 3.9 – Normalised covariance vs. distance for UMTS. ............................................................... 84 Figure 3.10 – Semivariance vs. distance for UMTS. ............................................................................ 84 Figure 3.11 – Cross-correlation vs. SNR difference for DCS. .............................................................. 86 Figure 3.12 – Covariance vs. SNR difference for DCS. ....................................................................... 86 Figure 3.13 – Normalised covariance vs. SNR difference for DCS...................................................... 87 Figure 3.14 – Cross-correlation vs. SNR difference for UMTS............................................................ 87 Figure 3.15 – Covariance vs. SNR difference for UMTS. .................................................................... 88 Figure 3.16 – Normalised covariance vs. SNR difference for UMTS. ................................................. 88 Figure 3.17 – Semivariance vs. SNR difference for UMTS.................................................................. 89 Figure 3.18 – General scheme of the measurement platform................................................................ 91 Figure 3.19 – Average power spectrum (averaged over more than 4800 2048-point FFTs) for some of the captured signals. Dashed lines represent the filter’s cut-off frequencies. ............................... 93 Figure 3.20 – Probability of detection versus SNR: (a) α = 0 dB (Pfa ≤ 0.01), (b) α = 0 dB (Pfa ≤ 0.10). ....................................................................................................................................................... 94 Figure 3.21 – Probability of detection versus SNR: (a) α = 1 dB (Pfa ≤ 0.10), (b) α = 2 dB (Pfa ≤ 0.10). ....................................................................................................................................................... 94 Figure 3.22 – Moving-averaged normalised power received for some captured signals (movingaveraging window of 100 samples)............................................................................................... 96 Figure 3.23 – Probability of detection versus sample length. ............................................................... 97 Figure 3.24 – Learning results tested on the channel 08A of the DAB-T standard. The curves compare the empirical ratio statistic to both the theoretical ratio distribution and Fisher-Snedecor distribution. ................................................................................................................................. 104 Figure 3.25 – Learning results tested on the channel 10A of the DAB-T standard. The curves compare the empirical ratio statistic to both the theoretical ratio distribution and Fisher-Snedecor distribution. ................................................................................................................................. 104 Figure 3.26 – Learning results tested on the channel 11B of the DAB-T standard. The curves compare the empirical ratio statistic to both the theoretical ratio distribution and Fisher-Snedecor distribution. ................................................................................................................................. 105 Figure 3.27 – Representation of a CA observing and accessing an RF environment. ....................... 108 Figure 3.28 – Percentage of time the UCB1 -based CA selects the optimal channel under various sensing errors frameworks (over 10 available channels)............................................................. 113 Figure 3.29 – UCB1 algorithm and Opportunistic Spectrum Access problem with sensing errors: regret simulation results. ............................................................................................................. 113 Reference DR9.3 6 / 126 216715 NEWCOM++ DR9.3 LIST OF TABLES Table 2.1 – JRRM traffic distribution variation scenarios. ................................................................... 15 Table 2.2 – JRRM VHOs strategies scenarios. ..................................................................................... 15 Table 2.3 – Service penetration per VNet. ............................................................................................ 21 Table 2-4 – Simulation scenarios. ......................................................................................................... 21 Table 2.5 – Network data and power efficiency for different systems and services. ............................ 27 Table 2.6 – Simulation parameters........................................................................................................ 33 Table 2.7 – Simulation parameters........................................................................................................ 50 Table 2.8 – Performance comparison of selected algorithms with 200 users distributed in observation area – Max Rate optimization ....................................................................................................... 51 Table 2.9 – Performance comparison of selected algorithms with 400 users distributed in observation area – Max Rate optimization ....................................................................................................... 52 Table 2.10 – Performance comparison of selected algorithms with 200 users distributed in observation area – Proportional Fair optimization............................................................................................ 52 Table 2.11 – Performance comparison of selected algorithms with 400 users distributed in observation area – Proportional Fair optimization............................................................................................ 53 Table 2.12 – Achieved rates for channel definition (2.67).................................................................... 72 Table 2.13 – Achieved rates for channel definition (2.68).................................................................... 72 Table 3.1 – Channels measured in this study: Analogical/digital tv, TErrestrial Trunked RAdio (TETRA), Terrestrial Digital Audio Broadcasting (DAB-T), Extended Global System for Mobile communications 900 downlink (E-GSM 900 DL), Digital Cellular System 1800 downlink (DCS 1800 DL) and Universal Mobile Telecommunications System Frequency-Division Duplex downlink (UMTS FDD DL).......................................................................................................... 92 Reference DR9.3 7 / 126 216715 NEWCOM++ DR9.3 LIST OF ACRONYMS 3G 3rd Generation ADC Analogical to Digital Conversor AMC Adaptive Modulation and Coding ASM Advanced Spectrum Management ATF Adaptive Throughput-Based Fairness AWGN Additive White Gaussian Noise BB Base Band BE Best Effort BS Base Station CA Cognitive Agent CDMA Code Division Multiple Access CFI Cell Fairness Index CoMP COordinated MultiPoint CR Cognitive Radio CRRM Common Radio Resource Management CSI Channel State Information CVRRM Cooperative VNet RRM DAB-T Terrestrial Digital Audio Broadcasting DAC Digital to Analogical Conversion DCS Digital Cellular System DL DownLink DSA Dynamic Spectrum Access DSAN Dynamic Spectrum Access Network EE Energy Efficiency EM Expectaion Maximization EvD Eigenvalue Decomposition FDD Frequency Division Duplex FDMA Frequency Division Multiple Access FPGA Field Programmable Gate Array FSRM Fairness-Based Sum Rate Maximization FSRM-P Fairness-Based Sum Rate Maximization with Proportional Rate Constraints GMM Gaussian Mixture Model GSM Global System for Mobile communications HHO Horizontal HandOver Reference DR9.3 8 / 126 216715 NEWCOM++ DR9.3 HOL Head Of Line ICIC Inter-Cell Interference Coordination IF Intermediate Frequency ILP Integer Linear Programming InP Infrastructure Providers IWF Iterative Water-Filling JRRM Joint Radio Resource Management LP Linear Programming LTE Long Term Evolution MAB Multi Armed Bandit MAI Multiple Access Interference ME Mobile Equipment MIMO Multiple Input Multiple Output ML Maximum-Likelihood MM Moment Method MMF Max-Min Fairness MMR Max-Min Rate MMSE Minimum Mean Squared Error MR Max Rate MRRM Multi Radio Resource Management MT Mobile Terminal NP-ED Neyman-Pearson Energy Detector NRT Non Real Time NSM Network Simplex Method OFDM Orthogonal Frequency Division Multiplexing OFDMA Orthogonal Frequency Division Multiple Access OSA Opportunistic Spectrum Access PF Proportional Fairness PRB Physical Resource Block PSC Packet SCheduler PSD Power Spectral Density PU Primary User 2 PU RC Per-User Unitary Rate Control RA Resurce Allocator RAT Radio Access Technology RR Round Robin RRA Radio Resource Allocation Reference DR9.3 9 / 126 216715 NEWCOM++ DR9.3 RRM Radio Resource Management RT Real Time SDR Software Defined Radio SINR Signal-to-Interference-plus-Noise Ratio SISO Single Input Single Output SNR Signal to Noise Ratio SRM Sum Rate Maximization SRM-P Sum Rate Maximization with Proportional Rate Constraints SU Secondary User SUS Semiorthogonal User Selection SVD Singular Value Decomposition TDMA Time Division Multiple Access TETRA TErrestrial Trunked RAdio TSD Transmit Selection Diversity TTI Transmission Time Interval UCB Upper Confidence Bound UL UpLink UMTS Universal Mobile Telecommunication System USB Universal Serial Bus USRP Universal Software Radio Peripheral VCG Vickey-Clarke-Groves VHO Vertical HandOver VLinks Virtual Links VNets Virtual Networks VNO Virtual Network Operators VNodes Virtual Nodes VNP VNet Providers VRRC VNet Requirements Radio Resource Control ZF Zero Forcing Reference DR9.3 10 / 126 216715 NEWCOM++ 1 INTRODUCTION 1.1 Objectives DR9.3 NEWCOM++ WPR9 addresses the development and evaluation of advanced Radio Resource Management (RRM) and spectrum management techniques for wireless communications systems in heterogeneous scenarios. In such scenarios different radio access technologies co-exist, so by a joint use of the available resources significant gains and a more efficient use of the spectrum can be achieved. Moreover, the developed strategies may be improved by introduction of cognitive network functionalities to provide the ability to adapt to changing conditions. Under this framework, the objective of this deliverable is to present the final report with the technical activities that have been carried out in the context of NEWCOM++ WPR9, with a particular focus on the ones realised during the third year, although proper references to the prior activities and preceding deliverable are given in order to build a self-contained deliverable. 1.2 Document structure This document is organised in accordance with the activities carried out in the different working groups (WG) of WPR9: • • Chapter 2 deals with RRM and JRRM algorithms at different levels, comprising the work of working groups WG1, WG2 and WG3 in the following subsections, after the introductory Section 2.1: o Section 2.2 focuses on the JRRM algorithms and their evaluation in heterogeneous networks, with the main focus on vertical handover (VHO) procedures o Section 2.3 addresses the Cooperative Radio Resource Management in Multi Virtual Networks, based on the network virtualisation concept where infrastructure is shared by virtual operators o Section 2.4 deals with the energy efficiency awareness in RRM o Section 2.5 deals with the spectrum allocation problem in a cellular OFDMA system, with the analysis both on single-cell and multi-cell level. o Section 2.6 addresses the application of game theory for optimization of the RRM in cognitive radio, focusing in particular on the definition of crystallized rate regions for MIMO transmission. Chapter 3 presents the outcomes of investigation within working groups WG4 and WG5 in the advanced spectrum management field. The results and proposed algorithms are given in two subsections following the introductory Section 3.1: o Section 3.2 describes the results of the measurement campaign to detect spectrum availability, with the aim to establish spatial and temporal usage patterns from primary users in different licensed bands, and also different spectrum sensing studies are presented. o Section 3.3 focuses on Dynamic Spectrum Access strategies for Cognitive Radio Networks, presenting on the one hand a joint learning detection framework, evaluated using the measurements of the campaign and on the other hand an opportunistic spectrum access mechanism with sensing errors.. Finally, in chapter 4 the main conclusions and the summary of integration activities within WPR9 are outlined. Reference DR9.3 11 / 126 216715 NEWCOM++ DR9.3 Reference DR9.3 12 / 126 216715 NEWCOM++ 2 2.1 DR9.3 RRM AND JRRM ALGORITHMS Introduction This chapter focuses on the work carried out inside WPR9 associated with the development and analysis of Radio Resource Management (RRM) strategies for heterogeneous wireless systems. Motivated by the clear trend towards the increase in the user demand for high bit rate data services including Internet accessibility everywhere and anytime, RRM functions targeting the most efficient use of the scarce radio resources in the different technologies have received increased attention. RRM strategies are responsible for taking decisions concerning the setting of different parameters influencing on the radio interface behaviour, including aspects such as the number of simultaneous users transmitting with their corresponding powers, transmission bit rates, the corresponding code sequences or sub-carriers assigned to them, the number of users that can be admitted in a given cell, etc. Furthermore, the future wireless arena is expected to be heterogeneous in nature, with a multiplicity of cellular, local area, metropolitan area and personal area technologies coexisting in the same geographical region. Also different deployments will be envisaged, with macro, micro, pico and femtocell deployments will be coexisting as the means to achieve the desired large capacities. Network heterogeneity has been in fact regarded as a new challenge to offer services to the users thanks to coordinating the available Radio Access Technologies (RATs), which on the other hand exhibit some degree of complementariness. In this way, not only the user can be served through the RAT that fits better to the terminal capabilities and service requirements, but also a more efficient use of the available radio resources can be achieved. This calls for the introduction of new RRM algorithms operating from a common perspective that take into account the overall amount of resources available in the existing RATs, and therefore are referred to as Common RRM (CRRM), Joint RRM (JRRM) or Multi RRM (MRRM). Under these considerations, one of the activities in this field in WPR9 that is presented in section 2.2 will address the JRRM strategies. In particular, the vertical handover (VHO) policies in a heterogeneous scenario with coexisting low and high bit rate networks in different configurations are studied, analysing the impact of traffic splitting policies in terms of different QoS indicators such as blocking probabilities or delay. Another concept that has recently received attention in the context of heterogeneous networks is the network virtualisation, in which different Virtual Network Operators share the network infrastructure. Such concept leads to the problem of how to perform an adequate RRM among the shared resources by the different operators. In that respect, section 2.3 will address the network virtualisation in heterogeneous environments where different RATs coexist. Section 2.4 will then focus on the introduction of energy efficiency awareness concepts on JRRM targeting the optimisation of the energy efficiency of the wireless networks by ensuring a proper match between real user data and required communication overhead. Different metrics are introduced to measure the power efficiency to transmit a certain useful data volume depending on the required overhead, and evaluated in a heterogeneous scenario with UMTS R99 and R5 technologies. Another trend in the wireless technology evolution has been the introduction in the new standards of more efficient technologies enabling the use of larger bit rates over a certain bandwidth. This is the case of the introduction of OFDMA and MIMO technologies in e.g. LTE or LTE Advanced standards. In such case, the need for efficient usage of the limited existing radio resources calls for the deployment of new strategies and revolutionary visions. Section 2.5 will in particular focus on the spectrum allocation in OFDMA networks, addressing first a single-cellular scenario and then the multi-cellular case, from the perspective of distributed schemes. The problem of Inter-Cell Interference Coordination (ICIC) in MIMO scenarios will also be considered within this section. Finally, section 2.6 will focus on the game-theoretic approaches to enable an optimisation of RRM procedures. In particular, a MIMO channel is considered and the crystallized rate regions for MIMO Reference DR9.3 13 / 126 216715 NEWCOM++ DR9.3 transmission are studied. Specific MIMO techniques are considered and the correlated equilibrium concept to the rate region problem is applied, proposing a new Vickey-Clarke-Groves (VCG) auction utility formulation and the modified regret-matching learning algorithm. 2.2 JRRM Strategies and Algorithms 2.2.1 Heterogeneous Networks VHO Model In heterogeneous networks it is not frequent to find simple analytical models in literature that are capable of extracting the main characteristics of a Vertical Handover (VHO) process and their impact into Quality of Service (QoS). For example in [1], [2], [3] and [4] relatively complex models are used to optimise and trigger VHOs, based on radio signal levels among others. However, a simple model that depends on geographical considerations is very useful to assess JRRM performance; in this way, based on simple considerations, the impact of several parameters at user and network levels on the JRRM QoS performance can be easily evaluated. Using a Horizontal Handover (HHO) analytical model as starting point, where user’s cell crossing probabilities, crossing rates, drops, blocking can be evaluated, [4], it is assumed that Radio Access Technology (RAT) types are grouped in clusters (see Figure 2.1) and can be modelled as a single cell (using cellular HHO characteristics). This approach enables the use of the previous model mechanism to extract some VHOs assumptions and theoretical results. This means that users’ previous transitions between cells borders of a given RAN are now translated to transitions between RATs borders (cluster based). This model is based on the traffic generation and on the heterogeneous network overall capacity. Thus, the model starts defining the number of active users, including the use of multiple services. The following conditions are assumed: network stability, infinite population, uniform users’ distribution in the area, Poisson and exponential distributions for service session arrival and duration processes, respectively. This proposed theoretical model was defined and proposed in the previous NEWCOM++ deliverable DR9.2 [5], where a reference scenario is proposed and some sub-scenarios variation impact are analysed. Then, relevant JRRM parameters variations are highlighted in this deliverable as described in the following sub-sections. Figure 2.1 – Different RAT clusters (reference scenario). 2.2.2 JRRM Theoretical Parameters’ Variation In order to present some possible results produced by the previous mentioned model, the most relevant input and output parameters need to be decided. Since the focus of this activity is JRRM performance, then, the overall JRRM QoS and VHOs related parameters in the model should be the most explored ones. In particular, nitial JRRM traffic distribution and VHO traffic percentage flow among RATs are the considered parameters to analyse the impact of their variation on the overall QoS. Reference DR9.3 14 / 126 216715 NEWCOM++ DR9.3 The model includes two JRRM parameters, PInit,r and PVHO, which aim to simulate the initial traffic distribution into a given RAT r , which is a JRRM influence policy, and the VHO policy influence among RATs, respectively. Taking into account the first one, it is possible to setup some distribution percentages over the generated traffic delivered to the correspondent RAT, enabling the QoS evaluation in each RAT. Table 2.1 presents two scenario variations over the reference one, Low Bitrate Networks (LBNs)-Centric and High Bitrate Networks (HBNs)-Centric [6]. These variations are useful to understand the impact of traffic distribution in each RAT QoS. The LBN (one single BS in cluster), will be the RAT type that will have more difficulties to observe high levels of traffic since compared with other RATs it provides low capacity and high coverage. Therefore, LBN related traffic percentage distributed by the JRRM entity should be low, to keep LBN QoS indicators under control. The second parameter PVHO proposed by the model, is related to VHO traffic transfer among RATs. Similar to previous cases, in Table 2.2 it is presented some traffic transfer variations among LBN, MBN and HBNs. Again, these traffic scenario variations take the same name as in the previous case. Basically the idea is to transfer traffic from one RAT to another; however, these transfer percentages have a physical limit, which is related to geographical superposition among RAT clusters. The VHO transfer rule, if negative, means that traffic flow will be the opposite with respect to the one symbolised by the arrow. Table 2.1 – JRRM traffic distribution variation scenarios. JRRM traffic distribution [%] RAT LBN – Centric Ref. HBN – Centric LBN 10 7 4 MBN 60 45 35 HBN 30 48 61 Table 2.2 – JRRM VHOs strategies scenarios. JRRM VHOs [%] VHO traffic transfer rule LBN - Centric Ref. HBN – Centric LBN→MBN -10 0 10 MBN→HBN -5 0 5 This sub-section presents the JRRM theoretical parameters’ variation results. These are presented observing simultaneously two dimensions, one dimension is the initial traffic and the second is the VHO policy percentage distribution variation/trend. These variations are performed according to Table 2.1 and Table 2.2 parameters, expecting to have significant and different impacts on RATs’ QoS indicators. These results are presented and discussed below. In Figure 2.2, it is possible to observe the LBN blocking probability variation. The worst case, from the LBN view perspective, is when initial traffic distribution and VHO policy push traffic to LBNs (LBN-Centric) in both axels, leading LBN blocking probability to about 13%. It is also possible to observe that initial traffic distribution has more influence when compared with the VHO variation policy. This is because the VHO percentage has less freedom on amplitude range. The best case is when initial distribution tends to HBN-Centric, in this situation LBN initial traffic percentage will move from 10 to 4% of the total generated traffic. Though this seams a small variation, with 4% LBN will handle less than 50% previous case (LBN-Centric), leading the LBN blocking probability to be less than 1%. Reference DR9.3 15 / 126 216715 NEWCOM++ DR9.3 LBN Blocking Probability [%] 14 12 10 8 6 HBN-Centric 4 Ref. VHOs Distribution 2 LBN-Centric 0 LBN-Centric Ref. HBN-Centric Initial Distribution Figure 2.2 – JRRM policies impact on LBN blocking. Anther QoS indicator is delay. Figure 2.3 presents results for LBNs and here, similar to PB, delay presents a parallel trend. However, note that delay values for LBN are only “acceptable” when HBNCentric case is plotted, this means that offered traffic to LBN should be less in order to reach the delay values set by services’ constraints. 700 LBN Delay [ms] 600 500 400 300 HBN-Centric 200 Ref. 100 VHOs Distribution LBN-Centric 0 LBN-Centric Ref. HBN-Centric Initial Distribution Figure 2.3 – JRRM policies impact on LBN delay. 2.3 Cooperative Radio Resource Management in Multi Virtual Networks In future virtual networks environments [7], [8], in which heterogeneous networks coexist and new business roles and models are expected. Virtual Networks (VNets) cooperation for RRM is one of the most relevant issues, in order to achieve an efficient integration of different wireless technologies and the maintenance of QoS requirements. The network virtualisation concept is based on network infrastructure sharing by different Virtual Network Operators (VNOs). Thus, RRM in VNets could be seen as a problem of wireless network sharing for multi-operator networks, which is already studied for the introduction of Mobile Virtual Networks Operators in 3rd Generation (3G) systems. Several RRM strategies for 3G multi-operator networks have been proposed in the literature, since there is a critical need for radio resource control among the multiple operators [9], [10], [11]. However, a considerable difference exists, since in that case operators are forced to use similar network functions, as defined by 3G specifications. Instead, in network virtualisation the existence of different types of functions and communication protocols for Reference DR9.3 16 / 126 216715 NEWCOM++ DR9.3 each VNet is a fundamental issue. Additionally, the scope of those works is limited to one wireless access technology, and cannot benefit from possible trunking gains obtained by the cooperation among different RATs. VNets pose some new challenges in the scope of cooperative RRM [12]. New stakeholders are expected in the market, like Infrastructure Providers (InP), VNet Providers (VNP) and VNOs. Since new relations and independencies must be considered, the interaction among these new stakeholders must be taken into account by cooperative RRM policies. Furthermore, the allocation of physical resources to different VNets introduces new constraints that should be considered to perform initial selection and handover decisions or radio resource allocation. At the RRM level, these constraints should also be taken in account, since the controlled radio resource unit pools are not static (from the operator view point). In fact, they are grouped according to the allocation of VNets, and may be reallocated to another VNet, or simply do not belong to any VNet. According to VNets’ environment characteristics, two different levels of RRM functions should be considered, Intra-VNet and Inter-VNet ones. The former allows managing how end-users of a VNet share the resources of that particular VNet; it is the VNO that can freely define what kind of RRM it uses within its VNet. The latter, from now on designated as Cooperative VNet RRM (CVRRM), is responsible for managing how physical resources are allocated to different VNets. CVRRM ensures that every VNet gets the amount of resources as negotiated in the VNet establishment phase. It should be stressed that it does not operate on the resources that are required by an individual end-user; instead, it considers the aggregated resource demands of different VNets, nevertheless, it can be triggered by individual demands that affect the aggregated ones. In a multi-access analogy, CVRRM, and also Intra-VNet RRM, are equivalent to the Multi-RRM [13] or Common RRM [14], with the difference of the operational context. Figure 2.4 – Inter-VNet RRM and Intra-VNet RRM. The CVRRM set of functionalities is devoted to the characteristics abstraction of heterogeneous wireless environments, from the virtualisation process, keeping the main cooperative RRM target, i.e., to optimise network resources usage and to provide the always best connectivity, while ensuring VNets QoS. The resources considered in the CVRRM context are the physical ones, nodes and links, and the virtual ones, Virtual Nodes (VNodes) and Virtual Links (VLinks). Radio resources, abstracted by channels, are also referred to in this scope. CVRRM strategies are based on a global knowledge of physical resources, their allocation to virtual networks, the resources neighbour mapping, and fundamental VNets characteristics to which resources are allocated. The VNet “owners” agreements (Inter-VNPs, Inter-InPs and VNOs) are also important information that should be known. Therefore, CVRRM can react not only to the changes in the amount of resources allocated to a VNet, but also to the end-user requests that affect the aggregated resources (VNodes/VLinks). Appropriate resources monitoring and evaluation can determine changes in the resource allocation in order to maintain or optimise VNet requirements. This evaluation is performed Reference DR9.3 17 / 126 216715 NEWCOM++ DR9.3 by a cost function that allows a unified comparison among all the resources, according to a given management policy, derived from [15]. Based on cooperative RRM concerns, namely, initial RAT access selection, vertical handover, and resources scheduling/allocation, three CVRRM main functions were identified: • VNet Requirements Radio Resource Control (VRRC) − that manages physical resources allocation to different VNets, in order to ensure the amount of resources negotiated at the VNet establishment; it takes the possible changes in capacity/availability of radio resources that affect VNet requirements into account, e.g., data rate, delay, jitter, packet loss, and error rates. • Initial VNet selection − allowing transparency to end-users in the process of VNet attachment and optimising VNets utilisation. • VNet Handover support − in order to guarantee the always best connectivity, even when the VNet coverage is impossible, therefore, allowing handover between different VNets. In order to establish a network model, some assumptions are taken: uniform coverage by all the wireless systems under analysis, and the inexistence of a specific requirement from the VNO related to the wireless technology in use. It is considered that VNOs do not care about the specific wireless technology being use, as long as the contractual requirements are ensured. Moreover, it is assumed that end-user nodes are mobile and multi-homed, i.e., capable of supporting different radio interfaces, so that it can connect to any available network. Concerning the wireless access technologies involved, one considers Time Division/Frequency Division Multiple Access (TD/FDMA), Code Division Multiple Access (CDMA), Orthogonal Frequency Division Multiplexing (OFDM), and Orthogonal Frequency Division Multiple Access (OFDMA), as they cover most of the current wireless systems (GSM, UMTS, WiFi, and LTE), which from now on are considered as examples of such access technologies. Although, radio channel multiple access definition for each wireless technology is different, a level of abstraction is added, enabling a common approach to manage all radio resources. It is considered that each wireless link is generically composed of channels, which varies in number and capacity according to the wireless technology involved. However, the characteristics of each technology are taken into account, in order to emphasise the specific factors that influence channel capacity. The main feature considered here is the channel data rate. VNets are classified according to their contractual requirements. CVRRM functions will interact with a Monitoring Entity (ME) (e.g., the In-Network Management resource monitoring in the context of the 4WARD project [16]), which provides real time measurements, like available resources quantity and quality, neighbouring resources and failure detection. Furthermore, it is assumed that a ME instance exists in the physical node, providing global resource monitoring information, and in each virtual node, collecting its own monitoring information. The monitoring of the whole VNet is done through the association of several MEs instantiated in each of its VNodes, constituting an aggregated system. It is assumed that the ME, mainly the real time monitoring part, monitors the wireless medium and the node, therefore, providing the cost function inputs to computation, in order to allow the comparison among resources, and among VNets. The strategies used in particular by VRRC are related to the contractual VNet requirements, and are reflected by key performance indicators weights in the cost function computation for resource evaluation. The VRRC algorithm uses monitoring information, to compare the actual capacity with the contractual one, then, deciding on radio resources (re)allocation to a given VNet. The selection of additional radio resources is made in two steps: • According to the radio resources availability into the same physical link. A VNet borrowing margin, similar to the one defined in [9], is associated to the VNet type and is adapted according to VNet usage. As an example, in a VNet with best effort requirements, channels Reference DR9.3 18 / 126 216715 NEWCOM++ DR9.3 may be transferred (borrowed) to perform the total amount of data rate required by a VNet with stringent requirements, if no other channels are available. The opposite is only possible if the VNet with stringent is running on low usage. • Based on the cost of neighbour resources, which reflects the resources availability according to an implemented strategy. The scanning time of this decision process is adapted dynamically depending on resources utilisation, variability of the radio interface, and VNets characteristics. VRRC may also decide on the migration or adaptation of the amount of resources allocated to the VNets, in order to optimise radio resource usage, e.g., when the virtual resource runs on low usage over a long period of time. In order to evaluate the VRRC performance, a set of output parameters was identified as follows: VNO satisfaction level, out of contract ratio and VLink utilisation. They are key indicators that allow a proper validation of the proposed model, by accessing critical issues related with the virtualisation process, such as virtual links with QoS guarantees. It is worthwhile to note that VNOs are indirectly the “users” from VRRC viewpoint. The VNO satisfaction level, SVNO, represents the VNO requests to use the remaining capacity, according to the contract established with the VNP, and this capacity is not available: SVNO = 1 − ( in RVL act RVL − 1) (2.1) where RinVL is the data rate offered to the VLink and RactVL is the actual VLink data rate. This equation is only applied when the data rate offered to the VLink is above the actual, and below the minimum contracted; otherwise, the value of SVNO is zero. SVNO accounts for the effective decrease in the amount of contracted resources perceived by the VNO. The analysis of this parameter should be made at VLink and VNet levels. Indirectly, it allows evaluating the network capability to react to radio/channel impairments that reduce the availability of the virtual resources. It can be used to monitor the virtual links, in order to detect contract violations. The out of contract ratio is defined as the period of time, over the total sampling one, for which VNet contracted capacity is not available. It is a global metric, independent of the service level experienced by the VNO, since in low VNe