Dynamic periodic location area update in mobile networks

Dynamic Periodic Location Area Update in

Mobile Networks

Yi-Bing Lin, Fellow, IEEE, Pei-Chun Lee, and Imrich Chlamtac, Fellow, IEEE

Abstract—In mobile communications networks, periodic lo-cation area update (PLAU) is utilized to detect the presence of a mobile station (MS). In 3GPP Technical Specifications 23.012 and 24.008, a fixed PLAU scheme was proposed for the Universal Mobile Telecommunications System (UMTS), where the interval between two PLAUs is of fixed length. We observe that MS presence can also be detected through call activities and normal location area update (NLAU). Therefore, we propose a dynamic PLAU scheme where the PLAU interval is dynamically adjusted based on the call traffic and NLAU rate. An analytic model is developed to investigate the performance of dynamic and fixed PLAU schemes. This paper provides guidelines to select parameters for dynamic PLAU.

Index Terms—Mobile communications network, mobility man-agement, periodic location area update, Universal Mobile Telecom-munications System (UMTS).

I. INTRODUCTION

M

OBILE communications networks have been evolved from the second generation (e.g., GSM) to the 2.5 generation (e.g., GPRS) and then to the third generation [e.g., Universal Mobile Telecommunications System (UMTS)] [7]. In this evolution, the concept of mobility management has remained the same. Consider the circuit-switched domain of UMTS [1], [2]. In order to track the mobile stations (MSs), the cells (the radio coverages of base stations) in UMTS service area are grouped into several location areas (LAs). To deliver services to an MS, the cells in the group covering the MS will page the MS to establish the radio link. To identify the LA of an MS, mobility management is required. In mobility management, the MS informs the network of its location through the LA update procedure. The update procedure is executed in two situations.

1) Normal location area update (NLAU) is performed when the location of an MS has been changed. Location change of an MS is detected as follows. The cells continuously broadcast their cell identities. The MS periodically listens to the broadcast cell identity and compares it with the cell identity stored in the MS’s buffer. If the comparison indi-cates that the location area has been changed, then the MS sends the location area update message to the network.

Manuscript received September 24, 2001; revised December 31, 2001, March 13, 2002, May 19, 2002, and June 13, 2002. The work of Y.-B. Lin was sup-ported in part by the MOE Program of Excellence Research under Contract 91-E-FA04-4, FarEastone, the Lee and MTI Center for Networking Research, NCTU.

Y.-B. Lin is with the Department of Computer Science and Information En-gineering, National Chiao Tung University, Hsinchu, Taiwan, R.O.C. (e-mail: liny@csie.nctu.edu.tw).

P.-C. Lee is with National Chiao Tung University, Hsinchu, Taiwan, R.O.C. I. Chlamtac is with the University of Texas at Dallas, Richardson, TX 75083 USA.

Digital Object Identifier 10.1109/TVT.2002.802984

2) Periodic location area update (PLAU) allows an MS to periodically report its “presence” to the network even if the MS does not move. A periodic LA update timer (PLAU timer or T3212 in [2]) is maintained in the MS. Corresponding to the PLAU timer, an implicit detach (ID) timer is maintained in the network. When the PLAU timer expires, the MS performs PLAU.

NLAU has been intensively investigated in the literature (see [3] and [7] and the references therein). Most studies on PLAU fo-cused on mobility database failure restoration [4], [6]. However, a major purpose of PLAU is to allow the network to detect if an MS is still attached to the network in the normal network oper-ation situoper-ation (i.e., the mobility databases do not fail). To our knowledge, this aspect has not been investigated in the literature. An important issue for PLAU is the selection of the period for the PLAU/ID timers. In 3GPP TS 23.012 [1] and TS 24.008 [2], the value is set/changed by the network and broadcast to every MS in the LA through the L3-RRC SYSTEM INFORMATION BLOCK 1 message on the broadcast control channel (BCCH). In this approach, the value is the same for all MSs in an LA. There are two issues regarding this fixed PLAU scheme.

1) How is the value determined when an MS first enters an LA?

2) Is it appropriate to have a fixed value during the MS’s stay in an LA?

This paper addresses the above two issues. As we will discuss in Section II, the value should be selected based on the call and movement activities of the MS. Therefore, we propose a dynamic PLAU scheme in this paper. We investigate the per-formance of this scheme and compare it with the fixed PLAU scheme. Our study provides guidelines to select parameters for dynamic PLAU.

II. DYNAMICPLAU SCHEME

Before we describe our solution for PLAU, we first introduces the concept of attach. In UMTS, the attach procedure allows an MS to be “known” by the network. For example, after the MS is powered on, the attach procedure must be executed before the MS can obtain access to the UMTS services. In a mobile com-munications network, four events can be utilized by the network to detect the presence of an MS attached to the network:

1) MS call origination (the MS makes an outgoing call); 2) MS call termination (the MS receives an incoming call); 3) PLAU;

4) NLAU.

Note that we consider NLAU as the fourth event although the main purpose of NLAU is to detect the movement of an MS.

When an MS is detached (disconnected) from the network due to abnormal reasons (e.g., battery removal, subscriber moving out of the service area, and so on), the MS will not originate a call. The network detects abnormal MS detach in one of the following two cases.

Case 1) The next PLAU occurs before the arrival of the next MS call termination. In this case, the network detects expiration of the ID timer and considers the MS de-tached. The next MS call terminations will not be delivered.

Case 2) The next MS call termination occurs before the ID timer expires. In this case, the network attempts to deliver the next incoming call to the MS but fails. After failure of the call setup, the network considers that the MS is detached and will disable future MS call terminations and PLAU timer.

In the call termination procedure, the network resources (trunks and so on) are reserved. In Case 2), these network resources are not released until the network detects that call setup fails. In other words, failed call setup wastes network resources, which should be avoided. To reduce the possibility of Case 2), one may shorten the interval of PLAU. On the other hand, short may result in large network signaling overhead. Therefore, should be carefully selected. A perfect PLAU mechanism will satisfy the following criteria.

Criterion 1: When the MS is attached, the presence of the MS is detected through call activities (either incoming or outgoing) or NLAUs, and PLAU is never performed. Criterion 2: When the MS is abnormally detached, the net-work detects the situation through a periodic update mech-anism [i.e., Case 1) holds], and failure call setup [i.e., Case 2)] never occurs.

If Criterion 1 is satisfied, the PLAU cost is zero when the MS is attached. If Criterion 2 is satisfied, then the network resources will not be wasted due to call setup failure. We show how to select the value with attempt to satisfy both Criteria 1 and 2. Suppose that the outgoing (MS originated) call arrival rate plus the NLAU rate to an MS is and the incoming (MS terminated) call arrival rate is . Then the net arrival rate is . Let

Statistics from mobile operators [7] indicate that 40% of the call activities are incoming calls to an MS. Therefore,

can be observed in a typical mobile network. When the MS is attached to the network, we expect to see a call (either incoming or outgoing) or an NLAU for every interval. If we select (where ) and after every call arrival is reset to , then there is a good chance that Criterion 1 is satisfied. Consider the example in Fig. 1(a) where calls or NLAUs arrive at , , and . For discussion purposes, we define presence checkpoint or checkpoint as the action to inform the network of the status of an MS (whether the MS is attached or not). We also define checkpoint event as an incoming call, an outgoing call, or an NLAU. Such an event results in checkpoint action. When the MS is attached, a checkpoint is triggered by an checkpoint event or PLAU. When the MS is abnormally detached, a checkpoint

Fig. 1. Examples where Criteria 1 and 2 are satisfied.

is triggered by expiration of ID timer or failure setup for MS call termination. In Fig. 1(a), a checkpoint occurs at and the PLAU timer is reset to (i.e., the next PLAU is expected to occur at time ). If is sufficiently large so that

, then the next checkpoint occurs at and the PLAU timer is reset to again. In this scenario, the PLAU is never performed and all checkpoints are triggered by call arrivals or NLAUs. If we select so that

(1) then there is good opportunity that Criterion 1 is satisfied. How-ever, if is too large, Criterion 2 is likely to be violated.

Fig. 1(b) illustrates a scenario when Criterion 2 is satisfied. In this figure, the MS is abnormally detached between two check-points. The previous checkpoint occurs at . The abnormal MS detach occurs at . After MS detach, the next PLAU oc-curs at . The next call termination occurs at . If , then the PLAU timer expires before the next call ter-mination arrives, and Criterion 2 is satisfied. Since the MS call termination rate is , to have a good chance to satisfy Criterion 2, we suggest that is selected such that

(2) From (1) and (2), if , it seems appropriate to select so that

(3) Based on the above discussion, we propose a dynamic PLAU scheme that dynamically selects according to the call and NLAU activities of an MS.

Dynamic PLAU Scheme

Step 0. Initially a default value is

Step 1. When a checkpoint event arrives, the following steps are executed in the network, specifically, the visitor loca-tion register (VLR) [7]:

Step 1.1. The interval between this

checkpoint event and the previous check-point event is computed and stored in

storage. The network stores the most

recent intercheckpoint event arrival time samples.

Step 1.2. The statistics are updated.

Step 1.3. Let be the intercheckpoint

event arrival time between the th

pre-vious checkpoint event and the 1st

pre-vious checkpoint event. The value is

computed as

(4)

where is selected following the

guide-line (3).

Step 1.4. The ID timer in the network is

reset with the value . The MS is

in-formed to reset its PLAU timer.

Step 2. When the network receives the PLAU message from the MS, the ID timer is reset

with the previously selected .

In the fixed PLAU scheme proposed in 3GPP TS 23.012 and TS 24.008, Step 2 is always executed, and Step 1 is never ex-ecuted. Also note that in the dynamic PLAU, the MS is in-formed to reset its PLAU timer by the network. In the standard GSM/UMTS procedures, when an MS requests for call origi-nation, NLAU, or PLAU, the network always acknowledges the request. The new value is included in GSM/UMTS acknowl-edgment messages issued by the network. In call termination, the network includes the new value in the call setup message. Therefore, no extra signaling messages are introduced by the dy-namic PLAU scheme at the cost that the acknowledgment and call setup messages are slightly modified. Note that in 3GPP TS 24.008, the value is broadcast to all MSs through the L3-RRC SYSTEM INFORMATION BLOCK 1 message on the BCCH, which cannot be used in our approach.

In a real GSM/UMTS network, the dynamic PLAU scheme can be implemented in the VLR as a microprocedure. This im-plementation can be vendor specific, which does not change any GSM/UMTS message flows. The message flows between the VLR and the MS follow the standard GSM/UMTS procedures.

III. ANALYTICMODELING

This section investigates the performance of dynamic PLAU and compares it with fixed PLAU proposed in 3GPP TS 23.012 and TS 24.008. Two performance measures are considered.

1) Let be the number of PLAUs occurring between two checkpoint events (incoming calls, outgoing calls, or NLAUs) when the MS is attached. The smaller the value, the lower the network signaling overhead caused by the PLAU mechanism. Criterion 1 is satisfied when

Fig. 2. The number of PLAUs between two checkpoint events when the MS is attached.

. Let and be the values for dynamic PLAU and fixed PLAU, respectively. We will derive the

expected values and .

2) Let be the probability that when an MS is abnormally detached, no failure call setup for mobile termination oc-curs [i.e., Case 1) holds]. It is clear that the bigger the value, the better the PLAU mechanism. Specifically, Cri-terion 2 is satisfied when . Let and be the values for dynamic PLAU and fixed PLAU, respectively. Telecommunications network operations suggest that incoming and outgoing call arrivals are Poisson streams [7], and the aggregate arrivals of the incoming and outgoing calls together with the NLAUs can be approximated as a Poisson stream. Following the above statement, we assume Poisson checkpoint event arrivals as in many other studies [4], [5], [8]. Therefore, in (4) are exponentially distributed, and has an Erlang-m density function with mean . That is

(5) The Laplace transform of the distribution is

(6) Fig. 2 shows the scenario when the MS is attached to the network and there are PLAUs between two checkpoint events

in dynamic PLAU. For , let .

Then, we have (7) shown at the bottom of the next page. For

(8) From (7), the expected number of is

Substitute (6) in (9) to yield

(10) Now we derive . Consider Fig. 2 again. This figure illustrates an example when the MS is attached to the network and there are PLAUs between two checkpoint events in fixed

PLAU. In this example, for , is a

fixed value, and the intercheckpoint event time is expressed as . For , the probability that is derived as

(11) For

(12) From (11), the expected number of is

(13) can also be derived from (10).

Consider the case when and . Equation (6) is rewritten as

(14)

From (6), (8), and (12), we have .

Also, from (8) and (14), we have

(15) Equation (15) is the same as (12). From (10) and (14), we have (16) Equation (16) is the same as (13).

Probability is derived as follows. Consider Fig. 3. Sup-pose that the previous checkpoint event or PLAU occurs at time , the abnormal MS detach occurs at , and the next PLAU occurs at . Define two events as follows.

Event A: No checkpoint event or PLAU occurs in period .

Event B: No MS call termination occurs in period .

It is apparent that Criterion 2 is satisfied if and only if event B occurs under the condition that event A occurs.

Fig. 3. Timing diagram for deriving and .

That is, . If the occurrence of abnormal MS detach is a random observer, then from residual life theorem and reverse residual life theorem [9], both and have the same density function

(17) Furthermore, events A and B are independent of each other and (18) Let be the probability that there is no MS termination occurring in a period . Since MS termination calls are a Poisson stream with rate , is expressed as [9]

(19) From (17)–(19)

(20) Now we derive . Consider Fig. 3 again. For fixed PLAU,

is a constant. Since occurrence of abnormal MS detach is a random observer, has a uniform distribution in interval [0,

]. From (19), is derived as

(21)

Fig. 4. The simulation flowchart.

We can also derive from (20) and (14)

The above analytic model is validated against the simulation experiments. We use a C program to implement the

simula-tion model that consists of three types of events: 1) Checkpoint (CKPNT); 2) PLAU; and 3) MS abnormal detach (AB_DE-TACH). The next CKPNT and AB_DETACH event arrival times are generated by the exponential random number generator, and all events are processed according to their timestamps. The sim-ulation flowchart is shown in Fig. 4. For a CKPNT event, steps 6–9 are executed. For a PLAU event, steps 10–11 are executed. For an AB_DETACH event, steps 12–16 are executed. In the simulation experiments, the abnormal MS detach occurs after an attach period exponentially distributed with mean that is

Fig. 5. Comparing analytic analysis with simulation experiments(m = 20).

times of an intercheckpoint arrival time interval. For , the simulation results are not sensitive to the values. In our simulation experiments, the confidence intervals of the 99% confidence levels are within 3% of the mean values in most cases. Fig. 5 shows that analytic analysis and simulation exper-iments are consistent for the values. The comparison results for other performance measures are similar and will not be pre-sented in this paper.

IV. NUMERICALEXAMPLES

Based on the analysis in Section III, we use numerical ex-amples to investigate the performance of dynamic PLAU and compare it with fixed PLAU.

By using (8), Fig. 6 plots as a function of . The figure indicates that even if we choose a small (e.g.,

or ), Criterion 1 can be satisfied with probability higher than 0.5. In this figure, the values are 0.667 13, 0.776 87, 0.850 43, and 0.899 74 for and respectively.

Based on (10), Fig. 7 plots against , where . The figure indicates that is a decreasing function of . When is small (i.e., ), if dynamic PLAU measures one more intercheckpoint arrival time sample (i.e,. is incremented by one), the performance is significantly improved. On the other hand, when is large , measuring more intercheckpoint arrival time sam-ples will not improve the performance. In Fig. 7, the values are 0.498 96, 0.287 22, 0.175 88, and 0.111 42 for

and respectively.

Based on (20), Fig. 8 plots against , where

and , and and . For ,

is not sensitive to the change of . Suppose that is selected. When ,

for . When , .

When , . Therefore, the

performance is significantly affected by (i.e., the frequency of incoming calls to the MS). As we mentioned before,

is observed in mobile network operations, and good performance can be expected. In this figure, the values are shown below. For , the values are 0.897 64, 0.863 94, and 0.831 94 for and respectively;

Fig. 6. ThePr[N = 0] performance.

Fig. 7. TheE[N ] performance.

Fig. 8. The performance.

for , the values are 0.809 01, 0.751 98, and 0.700 44

for and , respectively; for , the

values are 0.732 04, 0.659 37, and 0.596 65 for and respectively.

Note that the , , and values

perfor-Fig. 9. Relationship betweenE[N] and  (m = 20).

mance can only be achieved when the “optimal fixed values” are found. In reality, it is very difficult (if not possible) to guess such “optimal” values in advance. In Figs. 5–7, we demonstrate that by the adaptive mechanism, the dynamic PLAU scheme can achieve good performance close to the optimal fixed PLAU scheme.

The results in Figs. 7 and 8 indicate that and have conflicting goals. In other words, it is impossible to choose the values that minimize and maximize

at the same time. However, by choosing appropriate values, dynamic PLAU can satisfy both and restrictions, if such solutions do exist. For example, consider

. If the system requires that and

, then a value such as 1.1 satisfies the requirement

(with ).

Note that if the frequency of checkpoint events (call activities and NLAUs) changes from time to time, dynamic PLAU can automatically adapt to the change. Consider the scenario where for a long time (the first period) and then changes to 10 (the second period). Assume that the intervals for both the first and the second periods are the same. For dynamic PLAU, the period is adjusted as changes so that the value is kept as a constant (except for the short period for transition from to 10 ). On the other hand, the period is fixed in fixed PLAU. Therefore, in the first period and

in the second period. In other words, the value in the second period is ten times that in the first period. After the checkpoint rate changes, the fixed PLAU may only slightly improve the

performance at the cost of significantly degrading the performance. The consequence is that the average performance of fixed PLAU for the above mixed checkpoint traffic is worse than that for dynamic PLAU. Fig. 9 plots against for dynamic PLAU and fixed PLAU, where and

and . Note that is a function of , , and , while is a function of and . Therefore, when we choose a specific set of , the corresponding and values can be computed. Then we use these computed sets to plot this figure. The figure indicates that to achieve the same performance for the mixed checkpoint

traffic patterns, much less network signaling overhead for LA updates is expected in dynamic PLAU compared with that in fixed PLAU.

V. CONCLUSION

In mobile communications networks, periodic location area update is utilized to detect the presence of a mobile station. In 3GPP Technical Specifications 23.012 and 24.008, a fixed PLAU scheme was proposed for UMTS where the interval be-tween two PLAUs is of fixed length. We observe that MS pres-ence can also be detected through call and movement activi-ties. Therefore, we proposed a dynamic PLAU scheme where the PLAU interval is dynamically adjusted based on the call and NLAU traffic. An analytic model was developed to inves-tigate the performance of dynamic and fixed PLAU schemes. Our study indicates that compared with fixed PLAU, dynamic PLAU significantly reduces the network signaling traffic caused by periodic location area update.

As a final remark, in dynamic PLAU, storage and the mechanism maintaining or intercheckpoint arrival time samples for an MS can be practically implemented in the UMTS network (specifically, in the VLR). The value can be efficiently computed using the window averaging technique [7].

ACKNOWLEDGMENT

The authors would like to thank the three anonymous reviewers who have provided valuable comments that signifi-cantly improved the quality of this paper.

REFERENCES

[1] “3GPP 3rd Generation Partnership Project; Technical Specification Group Core Network; Location Management Procedures (Release 4),”, 3GPP TS 23.012 V4.0.0, 2001.

[2] “3GPP 3rd Generation Partnership Project; Technical Specification Group Core Network; Mobile Radio Interface Layer 3 Specification; Core Network Protocols—Stage 3 (Release 5),”, 3GPP TS 24.008 V5.0.0, 2001.

[3] I. F. Akyildiz, J. McNair, J. S. M. Ho, H. Uzunalioglu, and W. Wang, “Mobility management in next generation wireless systems,” Proc. IEEE, vol. 87, no. 8, pp. 1347–1385, 1999.

[4] Y. Fang, I. Chlamtac, and H. Fei, “Analytical results for optimal choice of location update interval for mobility database failure restoration in PCS networks,” IEEE Trans. Parallel Distrib. Syst., vol. 11, pp. 615–624, 2000.

[5] D. Hong and S. S. Rappaport, “Traffic model and performance analysis for cellular mobile radio telephone systems with prioritized and nonpro-tection handoff procedure,” IEEE Trans. Veh. Technol., vol. VT-35, pp. 77–92, Aug. 1986.

[6] Y.-B. Lin, “Failure restoration of mobility databases for personal com-munication networks,” ACM-Baltzer J. Wireless Networks, vol. 1, pp. 365–372, 1995.

[7] Y.-B. Lin and I. Chlamtac, Wireless and Mobile Network Architec-tures. New York: Wiley, 2001.

[8] S. S. Rappaport, “Blocking, hand-off and traffic performance for cellular communication systems with mixed platforms,” Proc. Inst. Elect. Eng. I, vol. 140, no. 5, pp. 389–401, 1993.

Yi-Bing Lin (M’95–SM’95–F’03) received the B.S.E.E. degree from National

Cheng Kung University, Taiwan, R.O.C., in 1983 and the Ph.D. degree in com-puter science from the University of Washington, Seattle, in 1990.

From 1990 to 1995, he was with the Applied Research Area at Bell Commu-nications Research (Bellcore), Morristown, NJ. In 1995, he became a Professor in the Department of Computer Science and Information Engineering (CSIE), National Chiao Tung University (NCTU), Taiwan, R.O.C. In 1996, he became Deputy Director of the Microelectronics and Information Systems Research Center, NCTU. During 1997–1999, he became Chairman of CSIE, NCTU. His current research interests include design and analysis of personal communica-tions services network, mobile computing, distributed simulation, and perfor-mance modeling. He is an Adjunct Research Fellow of Academia Sinica. He is an editor of Computer Networks, an Area Editor of ACM Mobile Computing and Communication Review, a Columnist of ACM Simulation Digest, and an Editor of the International Journal of Communications Systems, ACM/Baltzer Wire-less Networks, Computer Simulation Modeling and Analysis, and the Journal of Information Science and Engineering. He was Program Chair for the 8th Work-shop on Distributed and Parallel Simulation, General Chair for the 9th WorkWork-shop on Distributed and Parallel Simulation, Program Chair for the 2nd International Mobile Computing Conference, and Guest Editor for the ACM/Baltzer MONET Special Issue on Personal Communications.

Dr. Lin is an Associate Editor of IEEE NETWORK, an Editor of IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, an Associate Editor of IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, an Associate Editor of IEEE COMMUNICATIONS SURVEY AND TUTORIALS, and an Editor of IEEE PERSONALCOMMUNICATIONSMAGAZINE. He was a Guest Editor for the IEEE TRANSACTIONS ON COMPUTERS Special Issues on Mobile Computing and Wireless Internet, and a Guest Editor for IEEE COMMUNICATIONSMAGAZINE

Special Issue on Active, Programmable, and Mobile Code Networking. He is the author (with I. Chlamtac) of Wireless and Mobile Network Architecture (New York: Wiley, 2001). He received 1998 and 2000 Outstanding Research Awards from the National Science Council, R.O.C., and the 1998 Outstanding Youth Electrical Engineer Award from CIEE, R.O.C.

Pei-Chun Lee received the B.S.C.S.I.E. and M.S.C.S.I.E. degrees from

Na-tional Chiao Tung University (NCTU), Hsinchu, Taiwan, R.O.C., in 1998 and 2000, respectively, where she is currently pursuing the Ph.D. degree.

Her current research interests include design and analysis of a personal munications services network, computer telephony integration, mobile com-puting, and performance modeling.

Imrich Chlamtac (S’86–M’86–F’93) received the Ph.D. degree in computer

science from the University of Minnesota, Minneapolis.

Since 1997, he has been the Distinguished Chair in Telecommunications at the University of Texas at Dallas. He is the Sackler Professor at Tel Aviv Univer-sity, Israel, The Bruno Kessler Honorary Professor at the University of Trento, Italy, and University Professor at the Technical University of Budapest, Hun-gary. He has published close almost 300 papers in refereed journals and con-ferences and is the coauthor of the first textbook on Local Area Networks (Lex-ington Books, 1981) and of Mobile and Wireless Networks Protocols and Ser-vices (New York: Wiley, 2000), an IEEE NETWORKmagazine’s 2000 Editor’s Choice. He is the Founding Editor in Chief of ACM/URSI/Kluwer Wireless Net-works (WINET), ACM/Kluwer Mobile NetNet-works and Applications (MONET), and SPIE/Kluwer Optical Networks (ONM) Magazine.

Dr. Chlamtac is a Fellow of the ACM. He is a Fulbright Scholar and an IEEE Distinguished Lecturer. He received the 2001 ACM Sigmobile annual award and the IEEE ComSoc TCPC 2002 award for contributions to wireless and mobile networks and multiple best paper awards in wireless and optical networks.