The Design of power saving mechanisms in Ethernet Passive Optical Networks

1

The Design of power saving mechanisms in Ethernet Passive Optical Networks

Yun-Ting Chiang

Advisor: Prof Dr. Ho-Ting Wu

2014.01.09

2

Outline Introduction

Motivation and Research Objective Research Background

Passive Optical Network (PON) EPON (Ethernet PON)

Multi-Point Control Protocol (MPCP) Interleaved Polling with Adaptive Cycle Time (IPACT) Power Saving Mode on ONU

3

Outline

The Design of Power Saving mechanisms in EPON ONU power saving mode Power saving and MPCP The design of ONU three mode transform mechanism (Sleep to Doze) Transform mechanism and Timer

Scheduling for transform mechanism Downstream scheduling

Simulation result and Discussion Conclusion and Future work Reference

4

Introduction : Motivation and Research Objective

Motivation Greenhouse Effect End users require much bandwidth

Research Objective Consider both performance and power saving

5

Passive Optical Network (PON)

Optical line terminal (OLT) Optical network units (ONUs)

or Optical network terminals (ONTs)

Use broadcast on Downstream Use TDMA on Upstream

6

Multi-Point Control Protocol (MPCP) REPORT and GATE message REPORT

ONU to report its bandwidth requirements OLT passes REPORT to the DBA algorithm

GATE After executing DBA algorithm, OLT transmits GATE

downstream to issue up to four transmission grants to ONU Transmission start time Transmission length Timestamp (used by ONU for synchronization)

7

Interleaved Polling with Adaptive Cycle Time (IPACT)

OLT maintain a Table with Byte and RTT First grant, G(1), is set to some arbitrary value In polling cycle n, ONU measures its backlog in bytes

at end of current upstream data transmission & piggybacks the reported queue size, Q(n), at end of G(n)

Q(n) used by OLT to determine next grant G(n+1) => adaptive cycle time & dynamic bandwidth allocation

If Q(n)=0, OLT issues zero-byte grant to let ONU report its backlog for next grant

8Figure source: “IPACT: A Dynamic Protocol for an Ethernet PON”

9

Power Saving Mode on ONU : L. Shi’s research[1] Two energy-modes in ONU

In L. Shi, B. Mukherjee, and S. S. Lee, "Efficient PON with Sleep-Mode ONU: Progress, Challenges, and Solutions," refer two energy-modes including active and sleep modes. They separate high/low priority packet.

10

Power Saving Modeon ONU : L. Shi’s research[1] Two energy-modes in ONU

Time overhead and Power overhead When ONU switch mode will have overhead. Sleep to Active : Time overhead (2.125ms) Active to Sleep : Power overhead

Early wake up

11

Early wake up

ONU can receive GATE

msg

Because of Toverhead , ONU have wait 2.125ms to receive GATE msg. from OLT

12

ONU power saving mode

Active mode Tx : on, Rx: off

Doze mode Tx : off , Rx : on, support early wake up

Sleep mode Tx : off , Rx : off, support early wake up

p.s. Tx use to transmit REPORT Rx use to receive GATE

A

S

D

13

Power saving mode and MPCP

In MPCP, OLT and ONU have to handshake every 50ms. In order to decrease power consumption, we have to extend the rule.

14

Improve three energy-modes in ONU

Increase doze mode’s utilization

Decrease three energy-modes switching

15

The design of ONU three mode transform mechanism (Sleep to Doze)

A -> S

[1] No upstream and downstream data when OLT get ONUx’s REPORT.

A -> D

[2] No upstream data but has downstream data when OLT get ONUx’s REPORT

S -> A

[3] Upstream high priority data coming

// Early wake up

S -> D

[4] Stay at sleep mode for consecutive Y clock

// variable Y protects downstream high priority data ， Y is maximum of downstream high priority data delay.

16

The design of ONU three mode transform mechanism (Sleep to Doze)

D -> A

[5] Stay at doze mode for consecutive Z clock || upstream high priority data coming

// Timer avoids upstream long low priority data delay

// variable Y 、 Z protects upstream low priority data ， Y + Z is maximum upstream low priority data delay

p.s.

Active mode trigger: If report msg. request bandwidth = 0, means no upstream data.

17

Transform mechanism and Timer

Because we can’t use REPORT msg. in doze mode and sleep mode, we have to use timer to synchronize. Then OLT will know the state of ONUs.

18

Downstream scheduling

Although downstream slot and upstream slot are difference but there have some relationship.

// Downstream Data && GATE

Different from general EPON, because ONU[x] in sleep mode, OLT can’t send downstream data, downstream scheduling have to be considered.

ONUs’ doze mode maybe overlap so OLT need to select one of ONUs to send downstream data.

19

How to select an ONU to service We have 32 queue for downstream data to ONUs. Check downstream queues have data or not.Þ Ignore no data queues Separate Active table, Doze table and Sleep tableÞ Ignore sleep table If (Doze table is empty) {

If ( Active table is empty )

return false; // means no ONU have to service

else //

return an ONU which select by round-robin from Active table ;

} // end if

else //

return an ONU which select by round-robin from Doze table ;

20

Upstream scheduling

Using Limited service. Limited service : OLT grants requested number of

bytes, but no more than MTW OLT polling table record ONU state.

21

Simulation Result

ONU = 16 ONU queue size 100MByte EPON Frame size = 64Bytes ~ 1518 Bytes Channel capacity = 1Gbps Max rate = 100 * 1000 * 1000 = 100Mbps Guard time = 5 * 10-6

Y = 20ms, Timer, use to change mode from sleep to doze Z = 50ms, Timer, use to change mode from doze to active Simulation time 11s (1s warm-up)

22

Dynamic downstream loading

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

10

20

30

40

50

60

70

Utilization RatioUpstream loading = 0.5

Downstream High Priority Ratio= 50%Upstream High Priority Ratio= 50%

Active

Doze

Sleep

Downstream loading

Utilization R

atio

23

Dynamic downstream loading

0.050.10.150.20.250.30.350.40.450.50.550.60.650.70.750.80.850.90.95 102468

101214161820

Downstream High Priority DelayDownstream Low Priority Delay

Upstream loading = 0.5 Downstream High Priority Ratio= 50%

Upstream High Priority Ratio= 50%

D.H.DD.L.D

Downstream loading

Delay(m

s)

24

Dynamic downstream loading

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

1

2

3

4

5

6

Upstream High Priority DelayUpstream Low Priority Delay

Upstream loading = 0.5 Downstream High Priority Ratio= 50%

Upstream High Priority Ratio= 50%

U.H.DU.L.D

Downstream loading

Delay(m

s)

25

Dynamic downstream loading:upstream high priority ratio 100%

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

20

40

60

80

100

120

Utilization RatioUpstream loading = 0.7

Downstream High Priority Ratio= 50%Upstream High Priority Ratio= 100%

ActiveDozeSleep

Downstream loading

Utilization R

atio

26

Dynamic downstream loading:upstream high priority ratio 100%

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

2

4

6

8

10

12

14

16

18

20

Downstream High Priority DelayDownstream Low Priority Delay

Upstream loading = 0.7 Downstream High Priority Ratio= 50%Upstream High Priority Ratio= 100%

D.H.DD.L.D

Downstream loading

Delay(m

s)

27

Dynamic downstream loading:no upstream

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

20

40

60

80

100

120

Utilization RatioUpstream loading = 0

Downstream High Priority Ratio= 50%

Active

Doze

Sleep

Downstream loading

Utilization R

atio

28

Dynamic downstream loading:no upstream

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

2

4

6

8

10

12

14

16

18

20

Downstream High Priority DelayDownstream Low Priority Delay

Upstream loading = 0Downstream High Priority Ratio= 50%

D.H.DD.L.D

Downstream loading

Delay(m

s)

29

Dynamic upstream loading

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

20

40

60

80

100

120

Utilization RatioDownstream loading = 0.5

Downstream High Priority Ratio= 50%Upstream High Priority Ratio= 50%

Active

Doze

Sleep

Upstream loading

Utilization Ratio

30

Dynamic upstream loading

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Downstream High Priority DelayDownstream Low Priority Delay

Downstream loading = 0.5 Downstream High Priority Ratio= 50%

Upstream High Priority Ratio= 50%

D.H.D

D.L.D

Upstream loading

Delay(m

s)

31

Dynamic upstream loading

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 10

2

4

6

8

10

12

14

16

18

20

Upstream High Priority DelayUpstream Low Priority Delay

Downstream loading = 0.5 Downstream High Priority Ratio= 50%

Upstream High Priority Ratio= 50%

U.H.D

U.L.D

Upstream loading

Delay(m

s)

32

Conclusion

In this study, power saving mechanisms focus on reduce high priority downstream data delay in power saving EPON.

In order to raise up doze mode utilization, we design Sleep to Doze transform mechanism to increase it.

All results discuss between power saving and performance, it’s trade off. Maybe we can improve traffic scheduling or switching mechanism for future.

33

Future work

System parameter test Improve scheduling algorithm Doze to sleep transform mechanism Bidirectional transform mechanism

34

System parameter test

Timer : Y Timer : Z Different Timer set will change power consumption

and performance

35

Improve scheduling algorithm

We also can improve scheduling algorithm to reduce much power and increase performance

36

Doze to sleep transform mechanism

37

Bidirectional transform mechanism

38

Reference

[1] L. Shi et al., "Efficient PON with Sleep-Mode ONU: Progress, Challenges, and Solutions," IEEE Network Magazine, vol. 26, no. 2, pp. 36-41, march-april 2012.

[2] J. Mandin, "EPON Powersaving via Sleep Mode," IEEE 802.3av Meeting, 2008.

[3] L. Zhang et al., "Dual DEB-GPS Scheduler for Delay-Constraint Applications in Ethernet Passive Optical Networks," IECE Trans. Commun., Vols. E86-B, no. 5, pp. 1575-1584, 2003.

[4] Intel, "Intel and Ethernet," [Online]. Available: http://www.intel.com/content/www/us/en/standards/ethernet-innovation-case-study.html.

39

[5] COMMSCOPE, "White Paper GPON - EPON Comparison," 2012. [Online]. Available:http://www.commscope.com/. [Accessed 7 11 2013].

[6] J. Kani, S. Shimazu, N. Yoshimoto, and H. Hadama, "Energy efficient optical access networks - Issues and technologies," IEEE Commun. Mag., vol. 51, no. 2, pp. S22-S26, Feb. 2013.

[7] ITU-T, “G.Sup 45 : GPON power conservation,” 2009.

[8] D. Hood, Gigabit-capable Passive Optical Networks, Hoboken : John Wiley & Sons, 2012.

[9] ITU-T, Recommendation G.987.3, 2010-10.

40

[10] M. Ma, Y. Zhu, and T. Cheng, "A Bandwidth Guaranteed Polling MAC Protocol for Ethernet Passive Optical Network," Proc. IEEE INFOCOM, vol. 1, pp. 22-31, August 2003.

[11] S.-I. Choi, J-D Huh, "Dynamic Bandwidth Allocation Algorithm for Mutimedia Services over Ethernet PONs," ETRI Journal, vol. 24, no. 6, pp. 465-468, Dec. 2002.

[12] C. Assi et al., "Dynamic Bandwidth Allocation for Quality-of-Service Over Ethernet PONs," IEEE JSAC, vol. 21, no. 9, pp. 1467-77, Nov. 2003.

[13] G. Kramer, B. Mukherjee, and G. Pesavento, "IPACT: A Dynamic Protocol for an Ethernet PON (EPON)," IEEE Commun. Mag., vol. 40, no. 2, pp. 66-73, Feb. 2002.

[14] G. Kramer et al., "Supporting differentiated classes of service in Ethernet passive optical network," OSA J. Opt. Net., vol. 1, no. 8, pp. 280-298, Aug. 2002.

41

[15] R. Kubo, J. Kani, Y. Fujimoto, N. Yoshimoto, and K. Kumozaki, "Adaptive Power Saving Mechanism for 10 Gigabit Class PON Systems," IEICE Trans. Commun., Vols. E93-B, no. 2, pp. 280-288, Feb. 2010.

[16] X. Guo, G. Shou, Q. Xiao, Y. Hu, Z. Guo, “Toward green PON with adaptive sleep mode,” IC-NIDC, pp. 184-188, Sept. 2012.

[17] Y. Yan, S.-W. Wong, L. Valcarenghi, S.-H. Yen, D. Campelo, S. Yamashita, L. Kazovsky, and L. Dittmann, "Energy Management Mechanism For Ethernet Passive Optical Networks (EPONs)," May 2010.

42

Thanks for your listening