Presented by Jason L.Y. Lin Modelling and Performance Analysis of the Distributed Scheduler in IEEE802.16 Mesh Mode Min Cao, Dept. Electrical & Computer Engineering University of Illinois, Urbana-Champaign Wenchao Ma, Microsoft Research Asia Qian Zhang, Microsoft Research Asia Xiaodong Wang, Dept. Electrical Engineering Columbia University Wenwu Zhu, Intel China Research Center Proceedings of the 6th ACM international symposium on Mobile ad hoc networking and computing MobiHoc '05 Presented by Jason L.Y. Lin 2018/12/4 OPLab, Dept. of IM, NTU

Outline Introduction Background on IEEE 802.16 Mesh mode Modelling and Performance Analysis Simulation Results Conclusions and Future Work 2018/12/4 OPLab, Dept. of IM, NTU

Introduction (1/3) IEEE 802.16 MAC has two mode - point-to-multipoint (PMP) mode - multipoint-to-multipoint (mesh) mode In the mesh mode - nodes are organized in an ad-hoc fashion - there still be certain nodes that provide the BS function 2018/12/4 OPLab, Dept. of IM, NTU

Introduction (2/3) IEEE 802.16 has two mechanisms to schedule the data transmission in mesh mode - centralized scheduling - distributed scheduling In centralized scheduling - all the control and data packets need to go through the BS - the scheduling procedure is simple - but the connection setup delay is long 2018/12/4 OPLab, Dept. of IM, NTU

Introduction (3/3) In distributed scheduling - every node competes for channel access using a pseudo-random election algorithm based on the scheduling information of the two-hop neighbors - exhibits better flexibility and scalability - but distributed channel access control is more complex 2018/12/4 OPLab, Dept. of IM, NTU

Outline Introduction Background on IEEE 802.16 Mesh mode Modelling and Performance Analysis Simulation Results Conclusions and Future Work 2018/12/4 OPLab, Dept. of IM, NTU

Background on IEEE 802.16 Mesh mode the difference between 802.11 and 802.16 - 802.16 is a slotted system, and all transmissions must be synchronized - 802.16 uses a three-way handshaking to set up connection before data transmission - the control channel and data channel are separated in 802.16 - in 802.16, nodes can reserve multiple slots for the following packets without exchanging control message again 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (1/6) IEEE 802.16 distributed scheduling behavior - the control message and data packet are allocated in different time slots in a frame - there is no contention in the data time slots - employs a request/grant/confirm three-way handshaking procedure 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (2/6) 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (3/6) The Scheduling message, MSH-DSCH, contains the schedule and data subframe allocation information of the neighborhood. - NextXmtMx and XmtHoldoffExponent The transmission time for a station is an aggregate of some sequential transmission opportunities called eligibility interval The eligible interval length for a node is transmission opportunities. 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (4/6) After one eligibility interval, a station must hold off at least One station sets the first transmission slot after the holdoff time as the temporary next transmission opportunity 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (5/6) 2018/12/4 OPLab, Dept. of IM, NTU

IEEE 802.16 Distributed Scheduling Algorithm (6/6) Pseudo-random function mixing value : with the current node ID and the slot number as the inputs The channel contention result is correlated with the total node number, exponent value and network topology. Assumes the transmit time sequences of all the nodes in the control subframe form statistically independent renewal processes 2018/12/4 OPLab, Dept. of IM, NTU

Outline Introduction Background on IEEE 802.16 Mesh mode Modelling and Performance Analysis Simulation Results Conclusions and Future Work 2018/12/4 OPLab, Dept. of IM, NTU

Modelling and Performance Analysis Model and approach Two scenario - Collocated scenario * identical holdoff time * nonidentical holdoff exponents - General topology scenario Performance metrics estimation 2018/12/4 OPLab, Dept. of IM, NTU

Model and Approach (1/4) Assumptions (1) the counting process, , of each node eventually reaches its steady state and the intervals are i.i.d., that is, forms a stationary and ergodic renewal process (2) the renewal processes of different nodes are mutually independent at their steady states (3) when all the processes reach their steady states, we can assume that all the processes are initiated at t=-∞ and the time of renewal events of different processes are uncorrelated 2018/12/4 OPLab, Dept. of IM, NTU

Model and Approach (2/4) 2018/12/4 OPLab, Dept. of IM, NTU

Model and Approach (4/4) ：the expected number of competing nodes in slot s for the node of interest The probability that this node wins the slot is So the p.m.f. of S is 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (1/15) In collocated scenario - all nodes are one-hop neighbors of each other Identical Holdoff Exponent - assume equal holdoff exponents - hence when the node are collocated, the transmission interval has the same distribution 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (2/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (3/15) LEMMA 1. (Limiting Distribution of Excess Time) Let τ be the renewal interval, the limiting distribution of the excess time is for fixed , where and is an indicator function. By the stationary and ergodic assumption, 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (4/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (5/15) Figure 4:T he interval τ between two successive transmissions. 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (6/15) By the assumption that the renewal process is stationary and that the distributions of are identical, we can simply denote as . 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (7/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (8/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (9/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (10/15) denote as the number of nodes (among N-1 neighbors) which compete with node k in slot s. Denote as The expected number of nodes competing with node k in slot s is 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (11/15) The competing nodes in slot s for node k is Substituting (9) into (1) we get 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (12/15) we make a further approximation that 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (13/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (14/15) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Indentical Holdoff Exponent (15/15) Substitute and into (17) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (1/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (2/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (3/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (4/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (5/9) Assume that are geometrical distributed, that is, assume 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (6/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (7/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (8/9) 2018/12/4 OPLab, Dept. of IM, NTU

Collocated Scenario - Nondentical Holdoff Exponents (9/9) 2018/12/4 OPLab, Dept. of IM, NTU

General topology scenario (1/3) 2018/12/4 OPLab, Dept. of IM, NTU

General topology scenario (2/3) 2018/12/4 OPLab, Dept. of IM, NTU

General topology scenario (3/3) 4 1 3 3 2 4 1 2 1 2 3 2018/12/4 OPLab, Dept. of IM, NTU

Performance metrics estimation Let denote the time node A need to accomplish a three-way handshaking with node B. 2018/12/4 OPLab, Dept. of IM, NTU

Performance metrics estimation Assumptions - the renewal process of node A and B have run for a long time - follows a limiting distribution as the excess time when , we can assume that 2018/12/4 OPLab, Dept. of IM, NTU

Performance metrics estimation where α is a compromising factor. is a good choice for the identical holdoff exponent case and for the nonidentical holdoff exponents case 2018/12/4 OPLab, Dept. of IM, NTU

Performance metrics estimation 2018/12/4 OPLab, Dept. of IM, NTU

Outline Introduction Background on IEEE 802.16 Mesh mode Modelling and Performance Analysis Simulation Results Conclusions and Future Work 2018/12/4 OPLab, Dept. of IM, NTU

Simulation Results ns-2 simulator - network controller - scheduling controller - data channel component The set of possible exponent values is {0,1,2,3,4} 2018/12/4 OPLab, Dept. of IM, NTU

Transmission interval Figure 9:Simulation and analytical results on the expected transmission intervals for the identical exponent 2018/12/4 OPLab, Dept. of IM, NTU

Transmission interval Figure 10:Simulation and analytical results on the expected transmission intervals for the nonidentical exponent 2018/12/4 OPLab, Dept. of IM, NTU

Three-way Handshaking Time Figure 11:Simulation and analytical results on the three-way handshaking time for the identical exponent case 2018/12/4 OPLab, Dept. of IM, NTU

Three-way Handshaking Time Figure 12:Simulation and analytical results on the three-way handshaking time for the nonidentical exponent case with N = 10 2018/12/4 OPLab, Dept. of IM, NTU

Three-way Handshaking Time Figure 13:Simulation and analytical results on the three-way handshaking time for the nonidentical exponent case with N = 100 2018/12/4 OPLab, Dept. of IM, NTU

General Topology Scenario Table 1:Simulation and analytical results on the expected transmission intervals for the general topology 2018/12/4 OPLab, Dept. of IM, NTU

Outline Introduction Background on IEEE 802.16 Mesh mode Modelling and Performance Analysis Simulation Results Conclusions and Future Work 2018/12/4 OPLab, Dept. of IM, NTU

Conclusions and Future Work (1/2) The channel contention result is correlated with the total node number, exponent value and network topology. developed methods for estimating the distributions of the node transmission interval and connection setup delay also shed some light on the data subframe reservation scheme 2018/12/4 OPLab, Dept. of IM, NTU

Conclusions and Future Work (2/2) - to propose such a reservation scheme taking into account the tradeoff between system resource utilization and the connection QoS requirements 2018/12/4 OPLab, Dept. of IM, NTU

Thanks for your listening 2018/12/4 OPLab, Dept. of IM, NTU