QoS Scheduling for Heterogeneous Traffic in OFDMA-based Wireless Systems Youngki Kim Mobile R&D Laboratory KT, Korea Kyuho Son and Song Chong School of EECS KAIST, Korea IEEE GLOBECOM 2009 proceedings. Speaker:Tsung-Yin Lee
Outline Introduction Model Description and Problem Formulation Proposed QoS Scheduling Framework Simulation Result Conclusions
Introduction the key access technologies in current and next generation wireless systems is OFDMA Packet scheduling plays an important role in QoS provisioning by providing mechanisms for the resource allocation
Paper Goal provide QoS guarantee to the real-time traffic in multi-carrier wireless systems utility maximization of the non real-time traffic while providing QoS guarantee to the real-time traffic balance between QoS guarantee and utility maximization in a simple and organized manner
Model Description (1/2) Paper denote by S the set of all sub-channels in the system, NRT and NNRT, the set of all real-time (RT) and non real-time (NRT) flows RT flow, VoIP or MPEG, has its own QoS parameters such as maximum latency NRT flow has no explicit QoS parameters
Only Consider Downlink Model Description (2/2) Only Consider Downlink In the system, at each slot, the proposed scheduler determines the sub-channel assignment based on each flow’s current channel quality, minimum average throughput and individual packet deadline
Problem Formulation Proposed scheduling framework that maximizes the weighted sum rate of non real-time flows while maintaining QoS constraints of real-time flows in each time slot with the equal power allocation assumption
Problem Formulation (1/3) the derivative of utility function of flow i U’i(・) is used as a weight μij(t) is the achievable channel capacity when sub-channel j is assigned to flow i at time slot t
Problem Formulation (2/3) is the long-term throughput for flow i up to time slot t δij(τ) is the 0-1 indicator of allocating the sub-channel j to the flow i or not OFDMA constraint :
Problem Formulation (3/3) θi(t) is the actual amount of data allocated to real-time flow i at time slot t and πi(t) is given by : Mi is the minimum required average traffic rate of real-time flow i is the maximum possible data rate of real-time flow i at time slot t properly based on the newly introduced beta deadline parameter
Beta Deadline Parameter (1/2) urgent scheduling : which only considers the most urgent packets (required data rate is 6) strict priority scheduling : provide higher priority to the real-time traffic than non real-time traffic (required data rate is 18) paper may take a policy somewhere between these two extreme cases
Beta Deadline Parameter (2/2) lik is the length of the k-th packet of flow i eik is the time to expire value of the k-th packet of flow i Qi is the total number of packets of real-time flow i at time slot t
Real-time QoS Scheduling Paper can formulate the following maximum weighted bipartite matching (MWBM) problem to find the sub-channel allocation matrix : the number of sub-channels to be assigned to flow i at time slot t : the average sub-channel capacity of the flow i
Unweighted Bipartite Matching
Definitions Matching Free Vertex
Definitions Maximum Matching: matching with the largest number of edges
Definition Note that maximum matching is not unique.
Alternating Path Alternating between matching and non-matching edges. f g h i j d-h-e: alternating path a-f-b-h-d-i: alternating path starts and ends with free vertices f-b-h-e: not alternating path e-j: alternating path starts and ends with free vertices
Idea “Flip” augmenting path to get better matching Note: After flipping, the number of matched edges will increase by 1!
Idea of Algorithm Start with an arbitrary matching While we still can find an augmenting path Find the augmenting path P Flip the edges in P
Labelling Algorithm Start with arbitrary matching
Labelling Algorithm Pick a free vertex in the bottom
Labelling Algorithm Run Breadth-first search (BFS)
Labelling Algorithm Alternate unmatched/matched edges
Labelling Algorithm Until a augmenting path is found
Augmenting Tree
Flip!
Repeat Pick another free vertex in the bottom
Repeat Run BFS
Repeat Flip
Answer Since we cannot find any augmenting path, stop!
Weighted Bipartite Graph 3 4 6 6
Weighted Matching Score: 6+3+1=10 3 4 6 6
Maximum Weighted Matching Score: 6+1+1+1+4=13 3 4 6 6
Augmenting Path (change of definition) Any alternating path such that total score of unmatched edges > that of matched edges The score of the augmenting path is Score of unmatched edges – that of matched edges 3 4 6 6 Note: augmenting path need not start and end at free vertices!
Detailed Procedure the result of MWBM algorithm using average sub-channel capacity cannot give exact number of sub-channels to the flows
Non-Real-time QoS Scheduling general utility function is defined for α ≥ 0 α = 0 : maximum throughput α = 1 : proportional fairness α = ∞ : max-min fairness the minimum data rate that a dataflow achieves is maximized; secondly, the second lowest data rate that a dataflow achieves is maximized, etc
Simulation Environment VoIP traffic is based on G.711 codec standard and generates each VoIP packet every 20 ms, with 160-byte data Video streaming traffic has more bursty nature because packet size can be different according to the codec rate such as MPEG-FGS
Beta deadline parameter characteristics of VoIP traffic beta = 0 : strict priority beta = inf : urgent scheduling
Traffic class prioritization performance
Burst traffic response During the 2000 time slot and 3000 time slot, offered traffic rate increases up to 150% of the average traffic rate. During the 7000 time slot and 8000 time slot, offered traffic rate increases to 300% beta = 0 : strict priority beta = inf : urgent scheduling
Conclusions The proposed scheduling algorithm (beta deadline parameter) satisfies the QoS requirements of the real-time traffic and maximizes the utility of the non real-time traffic while utilizing the system resources efficiently