Open Access
Add to BibSonomyAbstract
We propose a novel Optical Network Units (ONU) migration mechanism within the Time and Wavelength Division Multiplexed PON (TWDM-PON) framework by rearranging the working ONUs to a minimum number of wavelengths and turning off the rest to save power. We show through simulation that the number of working wavelengths can be minimized up to a theoretical lower bound, e.g. 23%, under the typical ONU online profiles. We further investigate how the migration interval and delay influence the user Service Level Agreement (SLA). We find that under the example ONU online profiles, 99.99% ONU SLA can be realized with the migration delay of 1000 milli-seconds. However to realize 99.999% ONU SLA, the migration delay must be 100 milli-seconds or lower.
© 2013 Optical Society of America
1. Introduction
The global warming and climate change issues become extremely serious due to the rapidly increasing power consumption. It was reported that current networks consume 10,000 times more energy than the absolute minimum requirements [1]. The power saving has become an increasingly important concern in operators’ operating expenditures (OPEX) [2], among which access network equipment consumes a large share [3]. Reducing the power consumption in access networks is becoming imperative.
An obvious fact of access networks today is that the operators offer network access 24x7. Specifically, in Time Division Multiplexing Access (TDMA) platforms, it requires the Optical Line Terminal (OLTs) being online 24x7. However, the users connecting to the OLTs have unevenly distributed bandwidth demand during the daytime and night hours. Typically, residential users, for example, mostly power off their Optical Network Units (ONUs) or make them stay idle during the daytime. Conversely, they mostly have high bandwidth demand during the first half of night. The bandwidth demand for business users, on the contrary, is high during the daytime and low at night. Obviously, the 24x7 online mode wastes much power. The existing researches on power saving in access network are mostly focused on the ONU side by making the ONUs in doze or cyclic sleep according to the bursty nature of user’s traffic [3]. Meanwhile, the OLT keeps working 24x7.
In this paper, we are interested in reducing the power consumption at the OLT side, by taking into account the uneven distribution of the number of working ONUs at different time of the day. In particular, we are interested in Time and Wavelength Division Multiplexed PON (TWDM-PON) [4, 5], which has been accepted as the standard proposal in the International Telecommunication Union - Telecommunication Standardization (ITU-T) Sector. Compared with other candidates of next generation passive optical network stage 2 (NG-PON2), such as Wavelength Division Multiplexed PON (WDM-PON) [6] and Orthogonal Frequency Division Multiplexed PON (OFDM-PON) [7], TWDM-PON has better compatibility with the legacy time division multiplexing PON (TDM-PON) systems and is more cost effective.
TWDM-PON employs tunable transmitters and tunable receivers to meet the colorless requirement. This tunability has also been used to achieve traffic balance among the wavelengths [8]. However, it appears that, for most of the time, the wavelengths keep working regardless of the number of working ONUs. We propose a novel ONU migration mechanism to rearrange the working ONUs to a minimum number of wavelengths and turn off the rest ones. The migration agent inside the OLT constantly monitors the number of working ONUs and the wavelength utilization. We show through simulation that the number of working wavelengths can be minimized up to a theoretical lower bound, e.g. 23%, under the typical ONU online profiles. We further investigate how the migration interval and delay influence the user Service Level Agreement (SLA). We find that under the example ONU online profiles, 99.99% ONU SLA can be realized with the migration delay of 1000ms. However to realize 99.999% ONU SLA, the migration delay must be 100ms or lower.
The rest of this paper is organized as follows. Firstly, we give a short overview on the reported power saving techniques for PONs. Secondly, we introduce the developing TWDM-PON baseline architecture. Thirdly, we discuss the as-proposed ONU migration mechanism, and then evaluate the performance of the ONU migration mechanism with simulation. At last, conclusions are given in the section 5.
2. Related work in power saving for PON systems
Reducing the power consumption of access network is gaining more and more attention, as it typically consumes much more power than metro and core networks due to the huge number of distributed communication devices. In 2009, ITU-T published four basic power saving methods to reduce ONU power saving: power shedding, dozing, deep sleep and cyclic sleep [3]. In the literature, reported power saving methods for PON includes power shedding of non-essential parts [9, 10], slowing down access rate and putting some elements in sleep mode [11, 12], better integrated circuit design to maximize the sleep time [13, 14]. From the perspective of traffic flow and scheduling optimization, Yan et al. proposed to schedule the arrival of downstream traffic and the transmission of upstream traffic at the same time in each ONU to maximize the sleep time in a fixed bandwidth allocation scheme [15]. Dhaini et al. introduced a green bandwidth allocation framework by implementing a batch-mode transmission [16]. To maximize the sleep time, an adaptive power-saving mechanism was presented to dynamically control the ONU sleep time [17]. In a word, the power saving can be improved by maximizing the sleep time through dynamic control.
It is noted that the above-mentioned power saving methods focus only on the ONU sides. The OLTs run at full capacity 24x7 online. Reducing the power consumption of OLTs is as imperative from the network operators’ perspective. The Energy Efficient Ethernet (EEE) standardized as IEEE 802.3az is adopted for OLT Ethernet aggregators to suppress the power consumption of transmitter and receiver during sending and receiving idle frame [3]. Furthermore, wavelength routing technologies provide the solution of selective OLT sleep at low traffic [3].
The TWDM-PON standardized in ITU-T sector can reuse the above-mentioned power saving methods, which operate mainly on time domain. Considering the TWDM-PON is a composite of both time and wavelength domain, there are new power saving methods in wavelength domain. Today, researches are discussing the wavelength management and the wavelength assignment to achieve the load balance of the ONUs among the wavelengths [8]. All the wavelengths are working 24x7 online even though there are a few ONUs or none of ONU work at these wavelengths. Obviously that wavelength assignment is not power efficient. We find that it will become power efficient if the working ONUs are rearranged to a minimum number of wavelengths and the rest ones are powered off. The working ONUs are migrated from one wavelength to another during its working period. The migration trigger condition and threshold need to be defined in the migration mechanism carefully. Furthermore, migration interval and migration delay are taken into account to achieve the best power saving performance at OLT side and mitigate the service degradation of migrated ONUs.
3. ONU Migration in TWDM-PON
Typically, the TWDM-PON system supports multiple wavelengths pairs (e.g., wavelength pairs of {λ1, λM + 1}, {λ2, λM + 2} …) for down and upstream transmission, and each wavelength pair can support a subset of ONUs. As the ONUs served by one particular wavelength pair may go online and offline as time progresses, the number of working ONUs under that wavelength pair is dynamic. When the number of working ONUs under a wavelength pair (say, {λW + 1, λM + W + 1}) is small and there are enough timeslots unoccupied at other wavelength pairs (say, {λW, λM + W}), it would be reasonable to migrate these ONUs under {λW + 1, λM + W + 1} to {λW, λM + W}. As a result, the module and interface associated with {λW + 1, λM + W + 1} can be shutdown to reduce power consumption. Figure 1 shows the architecture of ONU migration in dynamic TWDM-PON system. Herewith M is the number of total wavelengths and W is the number of working wavelengths.
Clearly, the ONU migration is based on the tunability of receivers and transmitters, which is defined in the dynamic TWDM-PON standard. The receiver tunability is realized through a tunable filter. The tunability and wavelength management are also discussed in detail in [5, 8], and, what is more, we can reuse the information flows and messages of wavelength tunability report, wavelength assignment command, and wavelength change command defined in [8] to realize the ONU migration in this paper.
Figure 2(a) outlines the ONU migration procedure. The OLT continuously monitors bandwidth utilization and the number of working ONUs on each wavelength pair. If there is at least one wavelength that can be powered-off by migrating some of the working ONUs to other working wavelengths, OLT will send a migration command with information of the target wavelengths to the selected ONUs. It is possible that for load balancing purposes, the ONUs may be migrated to different wavelengths. This requires that the migration command be sent out on a one-by-one basis. Upon receiving the migration command, the ONUs execute the migration in two steps. In the first step, the affected ONUs deregister from the current wavelengths. In the second step, ONUs register on the target wavelength that is newly assigned. After completing the migration from current wavelength to the target wavelength, ONU sends a migration acknowledgment to the OLT. Once all ONUs complete migration, OLT will power off the wavelengths accordingly. Figure 2(b) illustrates how the migration and power-off of wavelength are controlled at OLT side.
It should be noted that the wavelength relation between ONUs and OLT is not fixed in the dynamic TWDM-PON framework. In the wavelength management framework in [8], all the wavelengths are working, and any ONU can register to any wavelength. However, in the ONU migration framework, only some of the wavelengths are working and we suggest the ONU register with a default wavelength pair, such as {λ1, λM+1} for downstream and upstream transmission. That is, this default wavelength keeps working all the time, no matter there are working ONUs or not. Upon receiving a registration request, the OLT admits the ONU if there is still available resource on the working wavelengths. Otherwise the OLT will power on one new wavelength to work. The OLT informs the ONU about the wavelength to be used. After receiving the wavelength information from the OLT, the ONU tunes its transmitter and receiver accordingly to complete the registration process.
The time interval with which the OLT performs migration has impact on migration and power saving performance. The number of working ONUs changes more rapidly, the shorter migration interval should be selected to achieve better power saving performance. Considering the fact that migrations happen only when the number of working ONUs decreases, without loss of generality, we select a period of time from T to T’, during which the number of working ONUs decreases monotonically. Figure 3 shows the number of working wavelengths at a few sample times with step size 15 minutes when selecting short and long migration intervals, e.g. 5 and 30 minutes, in contrast with the number of working wavelengths without migration.
It is apparent that the more frequent the OLT performs ONU migration, the more power saving it can achieve, especially when the number of working ONUs changes quickly with time. However, since migration operations inevitably lead to connection disruption on the ONUs, one may also want to minimize the number of ONU migrations in a given time. So selecting a proper migration interval/strategy is in fact a tradeoff between the desired power saving performance and the acceptable service availability. In the next section, we will show how the selection of migration interval in a periodic migration system will affect the power saving performance and the service availability. Our conclusions are that: when the number of working ONUs is more dynamic, changing the migration interval may result in more than 10% power saving difference, and one order of magnitude higher service availability. Different from the wavelength management in TWDM-PON proposed in [8], which focuses on balancing the traffic among a constant number of working wavelengths, the main purpose of ONU migration is to minimize the working wavelengths, and thus to save power consumption at OLT side.
4. Migration performance analysis
In this section, we investigate the power-saving performance of the proposed ONU migration mechanism with numerical simulation in Matlab. Typical values of the number of wavelength pairs in the current TWDM-PON shown in latest standards efforts are 4 and 8 [5]. But considering the rapid development of the PON technologies, the number of wavelength pairs is assumed to be 64 in our simulations [18]. The number of ONUs on one wavelength is further assumed to be 512 [5]. Thus the total number of ONUs under a TWDM-PON is 32,768. We assume that at different time of the day, each ONU is independently and identically online with probability defined by a normal distribution function N (μ, σ2). Without loss of generality, the distribution is assumed to be centered on 12 o’clock (μ = 12). The normal distribution function also approximates the number of online ONUs at different time of the day. We investigate three example ONU online profiles when the values of σ are 1/3, 2/3 and 3, respectively. Bigger values of σ imply more stable distribution of online ONUs against time, while smaller values mean that the number of ONUs online are more dependent on time and has a noticeable peak.
Figure 4(a) shows the three daily profiles of the number of working ONUs in a day. The three profiles have the same duty ratio of 23%, meaning that 23% ONUs are working on average in a day. This is also the theoretical lower bound for the number of working wavelengths. This duty ratio also determines the average percentage of closed wavelengths, as will be discussed in 4.2. In reality, the number of powered off ONUs may be less than 77%. But considering the fact that the number of idle ONUs occupies a considerable percentage of all ONUs during the daytime, we believe 23% is still a good approximation of the percentage of online ONUs.
4.1 The number of working wavelengths versus time of the day
Figure 4(b) shows the number of working wavelengths (W) versus time. It can be seen that the number of working wavelengths is closely related to the number of online ONUs (N). Ideally, it can be obtained by:
where C is defined as the maximum number of ONUs one wavelength can support. In reality, Eq. (1) provides a lower bound for the number of working wavelengths on an OLT. The more frequent an OLT performs user migration, the closer to the lower bound one might achieve in the number of working wavelengths. As will be seen shortly, although by migrating ONUs more often one can achieve better power saving performance, it also has adverse impact on ONU service availability. So selecting the migration interval is in fact a tradeoff between power saving performance and the migration cost.4.2 The power saving performance with different migration intervals
Figure 5(a) shows the percentage of closed wavelengths for different migration intervals. It shows that with migration intervals of 1 to 60 minutes, more than 70 percent of the wavelengths can be closed due to ONU migration. It also shows that the percentage of closed wavelengths decreases when the migration interval increases. This indicates that by migrating ONUs more frequently, more wavelengths can be closed and the working wavelengths can be used more efficiently. It can also be observed that with σ = 3 we get less significant power saving, as compared with σ = 1/3, and 2/3. As expected, the ONU migration shows higher percentage of closed wavelengths and better power saving performance for the ONU online profiles with smaller values of σ.
In Fig. 5(b) we compare the total number of ONU migrations for different ONU online profiles with various migration intervals. The trend of total number of ONU migrations with σ = 3 is flatter compared to σ = 1/3, and 2/3. It indicates that with smaller value of σ, the total number of ONU migrations is more dependent on the migration interval. The number with σ = 1/3 decreases more rapidly than σ = 2/3, and 3, when the migration interval increases. As discussed above, ONU migration can significantly improve the power efficiency at OLT by powering off some of the wavelengths. Figure 5 shows that longer migration interval results in less ONU migrations but also fewer number of closed wavelengths. In Fig. 6, we show the percentage of closed wavelengths normalized to the number of migrations under different migration intervals. It can be seen that longer migration intervals lead to higher unity gain.
4.3 Impact on the ONU Service Level Agreement (SLA)
As shown in Fig. 2(b), to perform migration, one ONU is first deregistered from the original wavelength (pair) then registered to another wavelength (pair), possibly interrupting the on-going services on the ONU. In this section, we analyze the migration delay in details and evaluate the potential negative impact on the ONU service level agreement (SLA), which is defined by:
where TM is the migration interval in seconds and S is the average number of migrations performed on each ONU in a day.To exam how the migration interval may affect the ONU SLA, we select three possible migration delay values 10ms, 100ms and 1000ms, which are all within the migration delay range. Figure 7 shows that the SLA performance is determined by the migration interval and migration delay.
The SLA performance gets better with longer migration interval for the ONU online profile with smaller values of σ, while for bigger values of σ, the performance doesn’t change too much with different migration intervals. In contrast, the performance depends more on the migration delay. When the average migration delay is less than 10ms, the ONU SLA can meet the 99.999% requirement. When the average migration delay is 100 ms, the ONU SLA is around 99.99% and when the migration delay is up to 1000ms, the ONU SLA will drop to 99.9% range. In general, the impact of ONU migration to the SLA can be minimized to 0.001% if the ONU migration can be completed within 10 ms.
5. Conclusion
In this paper, we introduce a novel ONU migration mechanism to improve the energy efficiency for the developing NG-PON2 standard. Through ONU migration, some of the wavelengths are powered off, saving the power consumption at OLT side. We analyze the power saving performance of the proposed ONU migration mechanism with typical ONU online profiles by calculating the percentage of closed wavelengths. The results show that the theoretical upper bound of percentage of closed wavelengths, e.g. 77% in our simulation, can be achieved based on the typical ONU online profiles. We find that the migration interval impacts the percentage of closed wavelengths when the number of working ONUs changes rapidly, while when the number of working ONUs is relatively stable, the migration interval doesn’t matter too much. The impact of migration delay to the SLA is also investigated in this paper. Our results show that 99.999% SLA can be guaranteed when the migration delay is controlled within 10ms.
Acknowledgments
This work is supported by NSFC (61271217, 61271216, and 61221001), Fok Ying Tung Education Foundation, 863 program (2011AA01A106, 2012AA011301), 973 program (2010CB328204-5), and Ministry of Education (20110073130006).
References and links
1. GreenTouch, http://www.greentouch.org
2. C. Lange, D. Kosiankowski, R. Weidmann, and A. Gladisch, “Energy Consumption of Telecommunication Networks and Related Improvement Options,” IEEE J. Sel. Top. Quantum Electron. 17(2), 285–295 (2011). [CrossRef]
3. J. Kani, “Power saving techniques and mechanisms for optical access networks systems,” J . Lightwave Technol. 31(4), 563–570 (2013). [CrossRef]
4. F. An, K. Kim, D. Gutierrez, S. Yam, E. Hu, K. Shrikhande, and L. Kazovsky, “SUCCESS: a next-generation hybrid WDM/TDM optical access network architecture,” J. Lightwave Technol. 22(11), 2557–2569 (2004). [CrossRef]
5. Y. Luo, X. Zhou, F. Effenberger, X. Yan, G. Peng, Y. Qian, and Y. Ma, “Time and Wavelength-Division Multiplexed Passive Optical Network (TWDM-PON) for Next-Generation PON Stage 2 (NG-PON2),” J. Lightwave Technol. 31(4), 587–593 (2013). [CrossRef]
6. G. Chang, A. Chowdhury, Z. Jia, H. Chien, M. Huang, J. Yu, and G. Ellinas, “Key Technologies of WDM-PON for Future Converged Optical Broadband Access Networks [Invited],” J. Opt. Commun. Netw. 1(4), C35–C50 (2009). [CrossRef]
7. C. Chow, C. Yeh, C. Wang, C. Wu, S. Chi, and C. Lin, “Studies of OFDM signal for broadband optical access networks,” IEEE J. Sel. Areas Comm. 28(6), 800–807 (2010). [CrossRef]
8. Y. Luo, S. Meng, and F. Effenberger, “Wavelength Management in Time and Wavelength Division Multiplexed Passive Optical Networks (TWDM-PONs)” in Proc. of IEEE Globecom, Anaheim, CA, 2971–2976 (2012). [CrossRef]
9. S. Wong, L. Valcarenghi, S. Yen, D. Campelo, S. Yamashita, and L. Kazovsky, “Sleep mode for energy saving PONs: advantages and drawbacks,” in Proc. of IEEE Globecom, Honolulu, HI, 1–6 (2009). [CrossRef]
10. J. Zhang and N. Ansari, “Toward Energy-Efficient 1G-EPON and 10GEPON with Sleep-Aware MAC Control and Scheduling,” IEEE Commun. Mag. 49(2), 33–38 (2011). [CrossRef]
11. R. Kubo, J. Kani, H. Ujikawa, T. Sakamoto, Y. Fujimoto, N. Yoshimoto, and H. Hadama, “Study and Demonstration of Sleep and Adaptive Link Rate Control Mechanisms for Energy Efficient 10G-EPON,” IEEE/OSA J. Opt. Commun. Netw. 2(9), 716–729 (2010). [CrossRef]
12. P. Chowdhury, M. Tornatore, and B. Mukherjee, “On the energy efficiency of mixed-line-rate networks,” in Proc. of OFC/NFOEC2010, paper OWY3. [CrossRef]
13. N. Suzuki, K. Kobiki, E. Igawa, and J. Nakagawa, “Dynamic Sleep-Mode ONU With Self-Sustained Fast-Lock CDR IC for Burst-Mode Power Saving in 10 G-EPON Systems,” IEEE Photon. Technol. Lett. 23(23), 1796–1798 (2011). [CrossRef]
14. S. Wong, S. Yen, P. Afshar, S. Yamashita, and L. Kazovsky, “Demonstration of energy conserving TDM-PON with sleep mode ONU using fast clock recovery circuit,” in Proc. of Opt. Fiber Commun. Conf./Nat. Fiber Opt. Eng. Conf., San Diego, CA (2010), paper OThW7. [CrossRef]
15. Y. Yan, S. Wong, L. Valcarenghi, S. Yen, D. R. Campelo, S. Yamashita, L. Kazovsky, and L. Dittmann, “Energy management mechanism for Ethernet passive optical networks (EPONs),” in Proc. of Int. Conf. on Communications, Cape Town, South Africa, 1–5 (2010). [CrossRef]
16. R. Dhaini, P. Ho, and G. Shen, “Toward green next-generation passive optical networks,” IEEE Commun. Mag. 49(11), 94–101 (2011). [CrossRef]
17. R. Kubo, J. Kani, Y. Fujimoto, N. Yoshimoto, and K. Kumozaki, “Adaptive power saving mechanism for 10 gigabit class PON systems,” IEICE Trans. Commun. E93-B(2), 280–288 (2010). [CrossRef]
18. S. Smolorz, H. Rohde, E. Gottwald, D. W. Smith, and A. Poustie, “Demonstration of a Coherent UDWDM-PON with Real-Time Processing,” in Proc. of NFOEC2011, paper PDPD4.
