1. Study on Power Saving for Cellular Digital Packet Data over a Random Error/Loss Channel Huei-Wen Ferng, Ph.D. Assistant Professor Department of Computer Science and Information Engineering (CSIE) Nation Taiwan University of Science and Technology (NTUST) Wireless Communications and Networking Engineering (WCANE) Lab E-mail: hwferng@mail.ntust.edu.tw

2. 6/23/2004 ICC 2004NTUST/WCANE Lab2 Outline Introduction Power Saving Mechanism (PSM) of CDPD Combination of SR-ARQ and PSM Numerical Examples and Discussions Conclusions

3. 6/23/2004 ICC 2004NTUST/WCANE Lab3 Introduction CDPD is a data network overlaid over AMPS. Elements of CDPD  M-ES, MDBS, MD-IS A mobile end system (M-ES) is the user end terminal. A mobile data base station (MDBS) works analogously to the base station (BS) in AMPS. Mobile data intermediate system (MD-IS) is responsible for mobility management and routing to other MD-ISs or fixed end systems (F- ESs).

4. 6/23/2004 ICC 2004NTUST/WCANE Lab4 Introduction The architecture of CDPD:

5. 6/23/2004 ICC 2004NTUST/WCANE Lab5 Power Saving Mechanism Two timers are maintained:  T203 and T204 timers T203 timer: element inactivity timer T204 timer: TEI notification timer An M-ES triggers a T203 timer when it completes a frame transmission.  The timer then counts down to zero.  The timer should be reset when data frames require the M-ES to receive or send before it expires.  Once the timer expires, the M-ES then enters the sleep mode.

6. 6/23/2004 ICC 2004NTUST/WCANE Lab6 Power Saving Mechanism To wake up the sleeping M-ES, T204 timer is maintained and periodically triggered by the MD-IS for each channel stream.  As T204 timer expires, the MD-IS broadcasts a frame containing TEIs with pending frames to be delivered but queued in their buffers at the MD-IS.  All M-ESs wake up to listen to the broadcast message during the broadcast period.

7. 6/23/2004 ICC 2004NTUST/WCANE Lab7 Power Saving Mechanism If the TEI for a specific M-ES is found in the frame, then the M-ES leaves the sleep mode.  The M-ES then sends a receiver ready (RR) frame to the MD-IS to notify the MD-IS that it is ready to receive the pending frames. If an RR frame is not received by the MD-IS, the MD-IS triggers the T204 timer again. If its own TEI of the M-ES is not found in the frame, then it remains in the sleep mode.

8. 6/23/2004 ICC 2004NTUST/WCANE Lab8 Power Saving Mechanism Listening Triggered

9. 6/23/2004 ICC 2004NTUST/WCANE Lab9 Combination of SR-ARQ and PSM Events and actions for the MD-IS:  1) Arrival of data frames from other MD-IS or network: Put these frames into the buffer if there is available space Discard overflowed frames. If the T203 timer for the M-ES does not expire, then deliver frames queued in the buffer with sequence numbers to the M-ES (via MDBS) and set a frame timer for each frame until the maintained sending window is greater than the window size.

10. 6/23/2004 ICC 2004NTUST/WCANE Lab10 Combination of SR-ARQ and PSM  2) Arrival of a data frame from an M-ES: Trigger the corresponding T203 timer. Check whether the frame is necessary to retransmit or not? If yes, send a NAK. Otherwise send an ACK and check whether the sequence number is within the receiving window?  If not, just discard the frame.  3) Timeout of T204 timer: Broadcast a TEI notification frame including all “sleeping” TEIs to all M-ESs. Trigger T204 timer again.

11. 6/23/2004 ICC 2004NTUST/WCANE Lab11 Combination of SR-ARQ and PSM  4) Timeout of T203 timer: Record the status of the corresponding TEI as “sleeping”.  5) Timeout of a frame timer: If T203 timer does not expire, then resend the frame and set the frame timer.  6) Receipt of RR: Trigger the corresponding T203 timer. Deliver frames queued in the buffer with sequence numbers to the M-ES (via MDBS) and set a frame timer for each frame until the maintained sending window is greater than the window size.

12. 6/23/2004 ICC 2004NTUST/WCANE Lab12 Combination of SR-ARQ and PSM  7) Receipt of an ACK or a NAK: Trigger the corresponding T203 timer. If an ACK is received, then remove all frames in the queue whose sequence numbers are smaller than the sequence number indicating by the ACK and update the sending window. If a NAK is received, then resend the expected frame. Finally deliver affordable frames in the buffer to the M-ES and set frame timers until the maintained sending window is greater than the window size.

13. 6/23/2004 ICC 2004NTUST/WCANE Lab13 Combination of SR-ARQ and PSM Events and actions for the M-ES:  1) Arrival of a data frame from the MD-IS: Check whether the frame is necessary to retransmit or not? If yes, send a NAK. Otherwise send an ACK and check whether the sequence number is within the receiving window?  If not, just discard the frame (a duplicated frame is received!). Finally trigger the T203 timer.

14. 6/23/2004 ICC 2004NTUST/WCANE Lab14 Combination of SR-ARQ and PSM  2) Time to wake up: Listen to the broadcast frame from the MD-IS. If its TEI is found in the frame, then send an RR frame to the MD-IS and trigger the T203 timer.  3) Timeout of T203 timer: The M-ES then enters the sleep mode.  4) Timeout of a frame timer: Resend the frame Set the frame timer and T203 timer.

15. 6/23/2004 ICC 2004NTUST/WCANE Lab15 Combination of SR-ARQ and PSM  5) Receipt of an ACK or a NAK: If an ACK is received, then remove all frames in the queue whose sequence numbers are smaller than the sequence number indicating by the ACK and update the sending window. If a NAK is received, then resend the expected frame. Finally deliver affordable frames in the buffer to the MD-IS and set frame timers and the T203 timer until the maintained sending window is greater than the window size.

16. 6/23/2004 ICC 2004NTUST/WCANE Lab16 Numerical Examples Simulation modeling and assumptions  One MD-IS and M-ES pair  Data frames for an M-ES are generated by a Poisson process (rate lambda) with uniform distributed number of frames (1~10)  Bit errors or frame losses are modeled by two independent Bernoulli random variables.  No error correcting code is employed.  Tft = 0.1 sec., T203 = 30 sec., T204 = 60 sec.  Buffer size is 10 or 30 times of data frames Performance measures  Dropping prob., sleeping ratio and waiting time (not shown in the slides)

17. 6/23/2004 ICC 2004NTUST/WCANE Lab17 Numerical Examples: Effect of SR-ARQ window size 1. BER=0.01, FLP=0, Buffer=10 2. SR-ARQ window size affects little to the PSM. 3. We then set window size to 4.

18. 6/23/2004 ICC 2004NTUST/WCANE Lab18 Numerical Examples: Effect of frame errors 1. BER>0.01 results in profound performance deviation 2. This says that the channel condition when BER>0.01 is indeed worse enough and retransmission turns into the main burden

19. 6/23/2004 ICC 2004NTUST/WCANE Lab19 Numerical Examples: Effect of frame errors 1. Increase of both BER and buffer size causes the sleeping ratio to drop down. 2. About 50% more sleeping ratio when BER=0.001 is gained as compared to BER=0.01.

20. 6/23/2004 ICC 2004NTUST/WCANE Lab20 Numerical Examples Effect of frame losses FLP<0.1

21. 6/23/2004 ICC 2004NTUST/WCANE Lab21 Conclusions System performance is insensitive to the SR-ARQ window size; hence, we suggest that 4 is suitable. Even with the SR-ARQ, the channel with bit error rate higher than 0.01 or the channel with frame loss probability larger than 0.1 deteriorates the system performance much. For BER <= 0.01 and FLP <= 0.1, the sleeping ratio of an M-ES is affected little by incorporating the SR-ARQ. Thus, SR-ARQ with proper parameters can recover frame errors or losses and retain lower power consumption for the typical channel condition, i.e., BER<=0.01 and FLP<=0.001.

22. 6/23/2004 ICC 2004NTUST/WCANE Lab22 Thank You!