A Utility-based Mechanism for Broadcast Recipient Maximization in WiMAX Multi-level Relay Networks Cheng-Hsien Lin, Jeng-Farn Lee, Jia-Hui Wan Department of Computer Science and Information Engineering, National Chung Cheng University, Taiwan IEEE Transactions on Vehicular Technology (IEEE TVT 2012) Outline Introduction Goal Network Model and Assumption Problem specification Multi-Level Utility-based Resource Allocation (ML-URA) Simulations Conclusions 2 Introduction The emergence of IEEE 802.16 WiMAX and advances in video coding technologies have made real-time applications possible. The granted applications (e.g., real-time IPTV Broadcast) Allocated limited time-slots (Resource Budget). 3 Problem This paper studies the resource allocation problem Broadcast receipt maximization in IEEE 802.16j IEEE 802.16j Multihop Relay Base Station(MR-BS) multiple Relay Stations(RSs) Mobile Stations(MSs) Broadcast data is sent by the MR-BS to a set of receivers How to allocate the given resource budget to maximize the number of MSs is a challenging issue. 4 Problem The broadcast receipt maximization problem 5 Problem The broadcast receipt maximization problem 6 Problem The broadcast receipt maximization problem 7 Problem The broadcast receipt maximization problem 8 Related works Existing researches heuristic resource allocation strategies single-level relay networks (two-hop relay networks) This paper models the resource allocation problem in IEEE 802.16j WiMAX multi-level relay networks (multi-hop) Multi-Level Broadcast Receipt Maximization (ML-BRM) problem 9 Goal To propose multi-level resource allocation mechanism Consider the multi-level relay paths and the required resource Maximize resource utilization in WiMAX multi-level relay networks 10 Network Model and Assumption In a WiMAX relay network, one MR-BS RS0 Each RS y (1 ≤ y ≤ Y) is denoted by RSy Y RSs N MSs that subscribe to a certain real-time program Each MS n (1 ≤ n ≤ N) is denoted by MSn This paper assumes that the real-time program, whose streaming data size is M Resource budget: rbudget total time slots in a TDD super frame 11 Network Model and Assumption The number of time slots required to transmit a broadcast stream varies MSs and RSs have different channel conditions MSs and RSs have different modulation schemes the transmission rates required for RSs to successfully send data also vary 12 Network Model and Assumption The transmission rate bx,y between sender x and receiver y based on one of the channel conditions, such as the SNR value sender x: MR-BS or RS receiver y: RS or MS The resource required by the receiver y: M/bx,y 13 Network Model and Assumption RAx: a node x with the allocated resource RAx all nodes whose required resource is not larger than RAx can receive the downlink data successfully through one downlink transmission from node x. MS RAx x MS MS 14 Network Model and Assumption For all RSs, the channel conditions are represented by R RS R1RS , R2RS ,..., RYRS where RyRS RRes y ,0 , RRes y ,1 ,..., RRes y ,Y records the resource required by RSy to receive streaming data from other RSs. RResy,y= 0: RSy doesn’t demand any resource from itself. R1RS RRes1,0 , RRes1,1 ,..., RRes1,8 RS5 R2RS RRes 2,0 , RRes 2,1 ,..., RRes 2,8 RS1 ... RS6 RS4 R8RS RRes8,0 , RRes8,1 ,..., RRes8,8 RS0 RS8 RS2 RS3 RS7 15 Network Model and Assumption Similarly, the matrix R MS R1MS , R2MS ,..., RNMS portrays the resource requirement of all MSs, where RnMS MResn,0 , MResn,1 ,..., MResn,Y records the resource that MSn requires to receive data from all RSs. R1MS MRes1,0 , MRes1,1 ,..., MRes1,8 MS1 MS2 R2MS MRes 2,0 , MRes 2,1 ,..., MRes 2,8 RS5 RS1 RS6 RS4 RS0 RS8 RS2 RS3 RS7 16 Network Model and Assumption Finally, the resource allocation vector is denoted by RA = [RA0, RA1, RA2, …, RAY ], where RAy represents the amount of the resource allocated to RSy. MS1 MS2 RS5 RS1 RS6 RS4 RS0 RS8 RS2 RS3 RS7 17 Network Model and Assumption U(): whether the MSn can receive data from RSy successfully. 1, if RAy MRes n , y 0 U ( RAy MRes n , y ) 0, otherwise RA1 = 5 MS2 RS0 U(RA1-MRes1,1) = U(5-3) = 1 U(RA1-MRes2,1) = U(5-7) = 0 MRes2,1 = 7 RS1 MS1 MRes1,1 = 3 18 Network Model and Assumption D(): whether RSy is eligible to receive real-time streaming data from the MR-BS when the current resource allocation RA is given. 1, if RAx RRes y , x and Dx ( RA) 1 Dy ( RA) 0, otherwise D0(RA) = 1: MR-BS is the source node of the real-time stream. RA1 = 5 RS3 RS0 D2(RA) = D2(5-3) = 1 D3(RA) = D3(5-7) = 0 RRes3,1 = 7 RS1 RS2 RRes2,1 = 3 19 Problem specification We now define the Multi-Level Broadcast Recipient Maximization (ML-BRM) problem. resource budget (rbudget) channel conditions of the wireless relay network (RMS and RRS ) ML-BRM searches for an allocation RA vector that will maximize the number of MSs receiving the real-time program. The ML-BRM problem is NP-complete 20 ML-URA Multi-Level Utility-based Resource Allocation Definition of Utility ui,y: the number of additional MSs divided by the extra resource that the network must allocate to the RSs on the relay path 21 ML-URA Construct single-source shortest path tree that is rooted at the MR-BS and connects all RSs. (SPy) ѱ(SPy) counts the number of RSs on SPy Γ(SPy, k) obtains the ID of the kth RS on SPy, 1 ≤ k ≤ ѱ(SPy) RS5 SP1 RS1 SP6 RS6 ѱ(SP = 2= 1 Γ(SP66), 1) Γ(SP6, 2) = 6 RS4 MR-BS RS8 RS2 RS3 RS7 22 ML-URA To derive the utility of a relay path ui,y count the number of additional MSs calculate the amount of extra resource required check if MSj can be served by SPy ……... ……... RSk RSk+1 RSy RS0 MSj Because of the broadcast nature of the wireless medium, MSj can receive data of the real-time program 23 ML-URA To derive the utility of a relay path ui,y count the number of additional MSs calculate the amount of extra resource required RSy is allocated MResi,y to serve MSi check if MSj can be served by RSy RSy MSi MSj Because of the broadcast nature of the wireless medium, MSj can receive data of the real-time program 24 ML-URA the union operation 1, if the above condition met FSP j , y FRS j , y FSP j , y FRS j , y 0, otherwise whether MSj has been served in previous rounds of the resource allocation process the additional number of MSs that can be served 25 ML-URA To derive the utility of a relay path ui,y count the number of additional MSs calculate the amount of extra resource required ……... ……... RSk RSk+1 RSy RS0 MSi 26 ML-URA To derive the utility of a relay path ui,y count the number of additional MSs calculate the amount of extra resource required RSk MSi Rsk+1 27 ML-URA The expression of the utility of a relay path ui,y is defined as follows: 28 ML-URA The ML-URA Mechanism Greedy procedure (ui,y) Find-Most-MS-Path procedure (number of MSs) 29 ML-URA_Greedy procedure stop conditions exists: (i) the entire resource budget has been allocated (ii) all MSs have been served. Greedy procedure 30 ML-URA_Greedy procedure Resource-Recycle procedure 31 ML-URA_Greedy procedure Two distinct paths that have the same utility value 2/2 5/5 32 ML-URA_Find-Most-MS-Path procedure Find-Most-MS-Path procedure 33 ML-URA_Find-Most-MS-Path procedure 34 Simulations 35 Simulations OPT => computes the optimal solution in a brute-force manner RAML 36 Simulations 37 Simulations 38 Conclusions The proposed ML-URA mechanism improve Resource utilization Performance 39