Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs)

Transcription

1 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Title: [Simulation Results for Final Proposal ] Date Submitted: [Aug 2013] Source: [Hongkun Li, Zhuo Chen, Chonggang Wang, Qing Li, Paul Russell Jr.] Company [InterDigital Communications Corporation] Address [781 Third Avenue, King of Prussia, PA , USA] Voice:[ ], FAX: [ ], Re: [Simulation Results for Final Proposal] Abstract: [This document presents simulation results on the MAC system design for (PAC)] Purpose: [To discuss performance of proposed system design for (PAC)] Notice: This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P Slide 1

2 Content 1. Performance of Discovery Procedure 2. Performance of Peering (Association) Procedure 3. Performance of Data Communication 4. References All the technique details in this presentation can be found in the final proposal [2]: IEEE Slide 2

3 1. Performance of Discovery Procedure Slide 3

4 Terms and Concepts--Discovery Peer Device (PD): A PAC device Tx PD: a PD that keeps sending discovery frames (i.e., beacon or repeated discovery request) within the proximity to be discovered Rx PD: a PD that is configured with a Tx PD, and keeps scanning the discovery frames to discover the desired Tx PD. To discover: A Rx PD scans discovery frames to find the desired Tx PD. To be discovered: A Tx PD sends out discovery frames to be discovered by a Rx PD within the proximity. Slide 4

5 Background--Discovery Discovery Schemes Beacon based discovery: A Tx PD sends beacon at beginning of its application frame to be discovered. Discovery request based discovery: A Tx PD sends to be discovered request once or multiple times after beacon in its application frame. Channel Management Scheme Contend for accessing the common channel (i.e. CCDCH) for channel allocation request. Insert application frame at the allocated location within a superframe. Topology Generation Follow the 2-step procedure in the TGD Drop Tx PD first, and then randomly drop Rx PDs within 50 meters of each Tx PD. Slide 5

6 Simulation Scenarios--Discovery Scenario 1: to discover scenario 1. All Tx PDs are turned on at time 0 and start contention based channel allocation request. 2. Then, all Tx PDs send beacon or to be discovered request on the allocated channel. 3. Then, all Rx PDs are randomly turned on. Scenario 2: to be discovered scenario 1. All Rx PDs are turned on at time Then, all Tx PDs are randomly turned on from time 0 and start contention based channel allocation request. 3. Then, all Tx PDs send beacon or to be discovered request on the allocated channel. Slide 6

7 Simulation Configuration Set 1--Discovery For the cases of 100, 500, 1000 PDs, there are 5, 10, 20 Tx PDs respectively. Parameter Value Slot size CCDCH length 1 ms Number of Superframes 100 Superframe length Simulation time Beacon interval Application frame length Bandwidth Channel data rate 10 ms (10 slots) 120 ms (120 slots) Number of Superframe * Superframe length * Slot size = 12 seconds 1 * Superframe length 5 ms (5 slots) 10 MHz 3 Mbps General parameters TGD revision 7 [1] Slide 7

8 Simulation Configuration Set 2--Discovery For the case of 5000 PDs, there are 50, 100, 200,,500, 1000 Tx PDs. Parameter Value Slot size CCDCH length 1 ms Number of Superframes 30 Superframe length Simulation time Beacon interval Application frame length Bandwidth 20 ms (20 slots) 320 ms (320 slots) Number of Superframe * Superframe length * Slot size = 9.6 seconds 1 * Superframe length 5 ms (5 slots) 10 MHz General parameters TGD revision 7 [1] Channel data rate 3 Mbps Slide 8

9 Discovery Performance Metrics Discovery latency to discover latency (measured in scenario 1): This metric is determined from the time that a Rx PD is turned on to the time that the Rx PD discovers the desired Tx PD. to be discovered latency (measured in scenario 2) This metric is determined from the time that a Tx PD is turned on for contention based channel request to the time that the Tx PD is successfully discovered by the first Rx PD. Power consumption to discover power consumption (measured in scenario 1): This metric is determined as the total power consumed by a Rx PD for listening to the channel and receiving either the beacon or discovery request message to discover a Tx PD. to be discovered power consumption (measured in scenario 2): This metric is determined as the total power consumed by a Tx PD from the time of requesting the channel to the time when all Rx PDs have discovered the Tx PD. Slide 9

10 Discovery Performance Metrics (Cont.) Rx PD discover ratio: This metric is determined as the ratio between the number of Rx PDs that successfully discover the desired Tx PD and the total number of Rx PDs Slide 10

11 Rx PD Discover Ratio Aug doc.: to discover Latency vs Ratio (Scenario 1) 1000 PDs, Beacon interval=superframe length=120ms 1 superframe length 0.4 5Tx, Beacon Based 10Tx, Beacon Based 20Tx, Beacon Based Tx, Discovery Req 10Tx, Discovery Req 5Tx, Discovery Req Latency (ms) All Rx PDs discover the desired Tx PDs within 1 superframe after turned on Slide 11 Discovery Request scheme achieves a shorter latency than the Beacon Based scheme due to: 1) Rx PD is randomly turned on 2) Repeated discovery requests offers more chances for discovery in the Discovery Request scheme

12 to discover Latency vs Ratio (Scenario 1) 5000 PDs, Beacon interval=superframe length=320ms superframe length All Rx PD discovers the desired Tx PD within 1 superframe after turned on Rx PD Discover Ratio Tx, Discovery Req 100Tx, Discovery Req Tx, Discovery Req 1000Tx, Discovery Req 1000Tx, Beacon Based Tx, Beacon Based 100Tx, Beacon Based 50Tx, Beacon Based Latency (ms) Slide 12

13 CDF of to discover Power Consumption (scenario 1) PDs, Beacon interval=superframe length=120ms CDF 0.4 5Tx, Discovery Req 10Tx, Discovery Req 20Tx, Discovery Req Tx, Beacon Based 10Tx, Beacon Based 5Tx, Beacon Based Power Consumption (mw*s) Slide 13

14 CDF of to discover Power Consumption (Scenario 1) PDs, Beacon interval=superframe length=320ms CDF 50Tx, Discovery Req 100Tx, Discovery Req Tx, Discovery Req 1000Tx, Discovery Req 1000Tx, Beacon Based Tx, Beacon Based 100Tx, Beacon Based 50Tx, Beacon Based Power Consumption (mw*s) Slide 14

15 Number of Tx PDs vs Average to discover Power Consumption (Scenario 1) 1.5 Average Power Consumption (mw*s) PD, Beacon Based 1000PD, Discovery Req 500PD, Beacon Based 500PD, Discovery Req 100PD, Beacon Based 100PD, Discovery Req Number of Tx PDs Slide 15

16 CDF of to be discovered Latency (Scenario 2) 1000 PDs, Beacon interval=superframe length=120ms 5Tx, Beacon Based 10Tx, Beacon Based 20Tx, Beacon Based 20Tx, Discovery Req 10Tx, Discovery Req 5Tx, Discovery Req Over 95% of Tx PD are discovered by all its Rx PDs within 3 Superframes from the starting time. CDF Listening period = 1 superframe channel allocaiton through contention = 1superframe Common channel at beginning of a superframe 1 superframe Latency (ms) Slide 16

17 CDF of to be discovered Power Consumption (Scenario 2) 1000 PDs, Beacon interval=superframe length=120ms CDF CDF 5Tx, Beacon Based Tx, Beacon Based 20Tx, Beacon Based Power Consumption (mw*s) Tx, Discovery Req Tx, Discovery Re 20Tx, Discovery Re Power Consumption (mw*s) Slide 17

18 Number of Tx PDs vs Average to be discovered Power Consumption (scenario 2) 6.4 Beacon interval=superframe length=120ms Average Power Consumption (mw*s) PDs, Discovery Req 500PDs, Discovery Req 100PDs, Discovery Req 1000PDs, Beacon Based 500PDs, Beacon Based 100PDs, Beacon Based Number of Tx PDs Slide 18

19 Discovery Latency Conclusion--Discovery The to discover latency will not exceed 1 Superframe length for all Rx PDs, which is independent of network density. The to be discovered latency will not exceed 4 Superframes for all Tx PDs, which is independent of network density. Discovery Request scheme has a shorter latency than Beacon Based scheme. Power Consumption Beacon Based scheme consumes similar amount of power as Discovery Request scheme. Rx PD discover ratio All Rx PDs are able to discover the desired Tx PD within 1 Superframe, i.e., the discovery ratio is 100%. Slide 19

20 2. Performance of Peering/Association Procedure Slide 20

21 Terms and Concepts Peering Requestor The PD that initiates the peering (association) process by sending a peering (association) request message to the Peering Responder. Peering Responder The PD that receives the peering (association) request message and sends a peering (association) response message to the Peering Requestor. CAP (Contention Access Period) First part of an application frame after application beacon, i.e., DCDCH. CFP (Contention Free Period) Second part of an application frame after CAP. Slide 21

22 Background--Peering (Association) Peering Schemes CAP/CFP-based Peering Requestors send peering request messages during CAP. Peering Responder sends response messages during CFP. Unicast separate responses to each peering requestor. Broadcast an aggregated response to all peering requestors. CAP-based Peering Requestors send peering request messages during CAP. Peering Responder sends response messages during CAP Unicast separate responses to each peering requestor. Broadcast an aggregated response to all peering requestors. Channel Access Schemes Fast Channel Access (FCA): A PD contends channel with a priority randomly chosen from [1, #Priority Classes] A PD performs backoff for a period randomly chosen from [1,t DCDCH ] when it senses channel busy or experiences a transmission failure. Slotted CSMA/CA Initial Backoff (IBF) PDs perform an initial backoff before contending for the channel. Slide 22

23 Fast Channel Access (FCA) Start fast DCDCH accessing Notes Parameters t ScanDCDCH : time window for scanning DCDCH t DCDCHi : initial waiting time before peer i detects the DCDCH again t EndDCDCH : End for current DCDCH t Peeri : Waiting time before detecting the DCDCH again by Peer i of P2PNW based on the priority class Inactive inject blocking signal DCDCH Slot1 Slot2 Slot3 No Receive APPframe beacon? Yes No Scan DCDCH for t ScanDCDCH Yes Is timed out t EndDCDCH? Peer 1 Beacon Detection t 1 Access DCDCH Peer 1 Data Peer 1 Data Is the channel occupied? No Yes Wait for random (t DCDCH ) time t DCDCH Peer 2 t 2 - Wait for t Peeri - Scan DCDCH Peer i t i (Peer i is timed out) DCDCH available DCDCH busy Priority-based DCDCH Access Is the channel occupied? No Access the DCDCH with association request Yes Slide 23 Fast Channel Access for Intra-P2PNW Communications through DCDCH

24 Aggregated Peering (Association) Response After received peering (association) request messages from PD A, B and C, responder Peer Z will broadcast a single aggregated peering (association) response message to three PD A, B and C. Slide 24

25 Simulation Configuration Network Configuration There are 100 PDs in the network (i.e. 99 Peering Requestors and 1 Peering Responder) Peering Requestors are randomly placed within 50 meters from the Peering Responder. All Peering Requestors start peering process at time 0. The transmission of each peering request or response message can be completed within 1 slot (i.e. 1 ms). Simulation Scenarios Set 1: CAP/CFP-based Peering, FCA priority range [1, 10], slotted CSMA/CA, IBF enabled or disabled. Set 2: CAP/CFP-based Peering, various FCA priority range, IBF disabled. Set 3: CAP/CFP-based & CAP-based Peering, FCA priority range [1, 10], slotted CSMA/CA, IBF disabled. Parameter Value Slot size DCDCH (CAP) length CFP length Application frame length Superframe length 1 ms 9 ms (9 slots) 1 ms (1 slots) 10 ms (10 slots) 100 ms (100 slots) General parameters TGD revision 7 [1] Slide 25

26 Performance Metrics--Peering (Association) Peering overhead: total number of messages (request and response) transmitted by all PDs until all Requestors are peered. Peering latency: the time (in Superframes) from all the Requestors start the peering procedure to the time that all Requestors are peered. Slide 26

27 Simulation Scenario--Set 1 Scenario Request message on Response message on Channel Access Priority Range IBF Aggregated Response 1 CAP CFP FCA [1,10] N Y 2 CAP CFP FCA [1,10] N N 3 CAP CFP FCA [1,10] Y Y 4 CAP CFP FCA [1,10] Y N 5 CAP CFP CSMA N/A N Y 6 CAP CFP CSMA N/A N N 7 CAP CFP CSMA N/A Y Y 8 CAP CFP CSMA N/A Y N Slide 27

28 Peering Overhead 1000 Aug doc.: Simulation Results--Peering Overhead Scenario 1: CFP IBF:0 Aggregated:1 Scenario 2: CFP IBF:0 Aggregated:0 Scenario 3: CFP IBF:1 Aggregated:1 Scenario 4: CFP IBF:1 Aggregated:0 Scenario 5: CSMA IBF:0 Aggregated:1 Scenario 6: CSMA IBF:0 Aggregated:0 Scenario 7: CSMA IBF:1 Aggregated:1 Scenario 8: CSMA IBF:1 Aggregated: t DCDCH in FCA or Backoff Window in CSMA (slots) Scenario 3 achieves the best performance ( i.e. FCA with Initial Backoff and aggregated response) The use of aggregated scheme reduces peering overhead. FCA performs better than CSMA. IBF reduces less overhead for FCA compared with CSMA. Slide 28

29 80 Simulation Results--Peering Latency Latency (superframes) FCA performs better than slotted CSMA/CA. IBF decreases the latency for slotted CSMA/CA, but has little impact on FCA. Scenario 1&2: CFP IBF: 0 Aggregated: 0 or 1 30 Scenario 3&4: CFP IBF: 1 Aggregated: 0 or 1 Scenario 4&5: CSMA IBF: 0 Aggregated: 0 or 1 Scenario 5&6: CSMA IBF: 1 Aggregated: 0 or t DCDCH in FCA or Backoff Window in CSMA (slots) Slide 29

30 Simulation Scenario Set 2 Scenario Request message on Response message on Channel Access Priority Range IBF Aggregated Response 1 CAP CFP FCA [1,10] N Y 2 CAP CFP FCA [1,10] N N 3 CAP CFP FCA [1,20] N Y 4 CAP CFP FCA [1,20] N N 5 CAP CFP FCA [1,50] N Y 6 CAP CFP FCA [1,50] N N 7 CAP CFP FCA [1,99] N Y 8 CAP CFP FCA [1,99] N N Slide 30

31 Peering Overhead Simulation Results--Peering Overhead Scenario 1: Prioirty Range:[1,10] Aggregated:1 Scenario 2: Prioirty Range:[1,10] Aggregated:0 Scenario 3: Prioirty Range:[1,20] Aggregated:1 Scenario 4: Prioirty Range:[1,20] Aggregated:0 Scenario 5: Prioirty Range:[1,50] Aggregated:1 Scenario 6: Prioirty Range:[1,50] Aggregated:0 Scenario 7: Prioirty Range:[1,99] Aggregated:1 Scenario 8: Prioirty Range:[1,99] Aggregated: t DCDCH (slots) Scenario 7 provides the best performance due to highest priority classes and aggregated response. The use of the aggregated scheme reduces peering overhead. Increasing number of priority classes reduces the peering overhead. Slide 31

32 Latency (superframes) Simulation Results--Peering Latency Scenario 1&2: Prioirty Range:[1,10] Aggregated:1 or 0 Scenario 3&4: Prioirty Range:[1,20] Aggregated:1 or 0 Scenario 5&6: Prioirty Range:[1,50] Aggregated:1 or 0 Scenario 7&8: Prioirty Range:[1,99] Aggregated:1 or 0 Scenarios 7&8 achieve the best performance due to highest priority classes. Increasing number of priority classes reduces the minimum peering latency of FCA. When t DCDCH becomes large, it dominates the peering latency over the other factors, such as priority class t DCDCH (slots) Slide 32

33 Simulation Scenarios--Set 3 Scenario Request message on Response message on Channel Access Priority Range IBF Aggregated Response 1 CAP CFP FCA [1,10] N Y 2 CAP CFP FCA [1,10] N N 3 CAP CFP CSMA N/A N Y 4 CAP CFP CSMA N/A N N 5 CAP CAP CSMA N/A N Y 6 CAP CAP CSMA N/A N N Slide 33

34 Peering Overhead 1000 Aug doc.: Simulation Results Peering Overhead Scenario 1: CFP FCA Aggregated:1 Scenario 2: CFP FCA Aggregated:0 Scenario 3: CFP CSMA Aggregated:1 Scenario 4: CFP CSMA Aggregated:0 Scenario 5: CAP CSMA Aggregated:1 Scenario 6: CAP CSMA Aggregated: t DCDCH in FCA or Backoff Window in CSMA (slots) Scenario 5 provides the best performance ( i.e. CAP/CFP based scheme with FCA and aggregated response). CAP/CFP-based peering scheme performs better than CAP only peering scheme. The use of aggregated scheme reduces the peering overhead for all schemes. Slide 34

35 Latency (superframe) Simulation Results--Peering Latency Scenario 1&2: CFP FCA Aggregated:1 or 0 Scenario 3&4: CFP CSMA Aggregated:1 or 0 Scenario 5: CAP CSMA Aggregated:1 Scenario 6: CAP CSMA Aggregated:0 Scenario 5&6 achieve the best performance ( i.e. CAP/CFP-based scheme with FCA). CAP/CFP-based peering scheme performs better than CAP only peering scheme. The use of aggregated scheme reduces the peering latency for CAP-based scheme t DCDCH in FCA or Backoff Window in CSMA (slots) Slide 35

36 Conclusions Aggregated peering response scheme reduces the peering overhead for both slotted CSMA/CA and FCA. CAP/CFP-based peering scheme performs better than CAP only peering scheme. Fast Channel Accessing (FCA) performs better than slotted CSMA/CA. Initial Backoff (IBF) reduces the peering overhead more for slotted CSMA/CA than FCA. Slide 36

37 3. Performance of Data Communication Slide 37

38 Terms and Concepts Tx PD: a PD that generates data packet and sends to the peered Rx PD. Rx PD: a PD that receives the data packet from its peered Tx PD. Slide 38

39 Background Data transmission is contention free within each application frame. Superframe Beacon 1 App Beacon1 (App1) App Beacon2 (App2) App Beacon3 (App3) Superframe Beacon 2 App Beacon1 (App1) App Beacon2 (App2) App Beacon3 (App3) CCDCH (1 st i slots) App1 DCDCH (1 st i 1 slots) App Frame 1 App2 DCDCH (1 st i 2 slots) App3 DCDCH (1 st i 3 slots) App Frame 2 App Frame3 App Frame 1 App Frame 2 Common Period Application Period Superframe1 Common Period Application Period Superframe2 Slide 39

40 Simulation Configuration--Data Communication Parameter Value Slot size 1 ms CCDCH length 20 ms (10ms for number of Tx PD < 50) Application frame length Number of application frames in a superframe 1 Beacon + 1 MPDU 30 (10 for number of Tx PD < 50) Superframe length (Number of application frames in a superframe * Application frame length) + CCDCH length Beacon interval Bandwidth 1 superframe length 10 MHz Channel data rate 9 Mbps (QPSK, 3/4) General parameters TGD revision 7 [1] Traffic model Full buffer & Poisson arrival Slide 40

41 Performance Metrics Area sum goodput: Mbps/km 2 Jain s fairness index MAC-to-MAC latency (only for Poisson arrival) From the time instant that the MAC at Tx PD decides to transmit a packet to the time instant that the Rx PD successfully receives the packet at MAC. Data packet reception efficiency (ratio) The total number of successfully received packet to the total number of transmitted packet including retransmission procedure. Slide 41

42 Area Sum Goodput (1) Aera sum goodput (Mps/km 2 ) MPDU size=512 bytes, Superframe length=30ms Poisson, IAT=100ms Poisson, IAT=10ms Poisson, IAT=1ms Full Buffer Poisson arrival process has the same performance as the Full Buffer model on the long term, if the IAT is smaller than Superframe length Number of Tx IAT: inter-arrival time Slide 42

43 Aera sum goodput (Mps/km 2 ) Aug doc.: Area Sum Goodput (2) Superframe length=80ms, MPDU size=512 bytes Poission Arrival, IAT=100ms Poission Arrival, IAT=10ms Full Buffer Poisson arrival process has the same performance as the Full Buffer model on the long term, if the IAT is smaller than Superframe length Number of Tx Slide 43

44 MAC-to-MAC latency (1) CDF Poisson arrival, MPDU size=512 bytes, superframe length=30ms Tx, IAT=100ms 10Tx, IAT=100ms 20Tx, IAT=100ms 20Tx, IAT=10ms 10Tx, IAT=10ms 5Tx, IAT=10ms The MAC-to-MAC latency of a data packet is bounded by Superframe length Latency (ms) Slide 44

45 CDF Aug doc.: MAC-to-MAC latency (2) Poisson Arrival, MPDU size=512 bytes, superframe length=80ms 50Tx, IAT=100ms 100Tx, IAT=100ms 200Tx, IAT=100ms 400Tx, IAT=100ms 512Tx, IAT=100ms 50Tx, IAT=10ms 100Tx, IAT=10ms 200Tx, IAT=10ms 400Tx, IAT=10ms 512Tx, IAT=10ms Latency (ms) Slide 45 The MAC-to-MAC latency of a data packet is bounded by Superframe length.

46 Fairness and Efficiency Jain s fairness index is always close to 1 due to: All Tx PDs have equal opportunity to send data within a superframe through the CFP of their application frames. Data packet reception efficiency (ratio) is always 1 due to: Packet error rate is 0 with channel model (i.e., path loss within 50 meters) and MCS (QPSK and ¾ coding rate). Data is transmitted over CFP within each application frame. Slide 46

47 Conclusion MAC-to-MAC latency is bounded by the Superframe length due to the contention free data transmission. Each Tx PD achieves almost the same throughput ( i.e., the fairness index is close to 1). Slide 47

48 References [1] IEEE Technical Guidance Document [2] Interdigital s final proposal: IEEE Slide 48

49 Thank You! Any Questions? Slide 49