BIP-Based Alarm Declaration and Clearing in SONET Networks Employing Automatic Protection Switching

Transcription

1 Vol:5, o:, 0 BI-Base Alarm Declaration an Clearing in SOET etworks Employing Automatic rotection Switching Vitalice K. Ouol an Cemal Aril International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 Abstract The paper examines the performance of bit-interleave parity (BI) methos in error rate monitoring, an in eclaration an clearing of alarms in those transport networks that employ automatic protection switching (AS). The BI-base error rate monitoring is attractive for its simplicity an ease of implementation. The BI-base results are compare with exact results an are foun to eclare the alarms too late, an to clear the alarms too early. It is conclue that the stanars evelopment an systems implementation shoul take into account the fact of early clearing an late eclaration of alarms. The winow parameters efining the etection an clearing threshols shoul be set so as to buil sufficient hysteresis into the system to ensure that BI-base implementations yiel acceptable performance results. Keywors Automatic protection switching, bit interleave parity, excessive bit error rate T I. ITRODUCTIO HIS paper examines the performance of bit-interleave parity (BI) methos in error rate monitoring, especially for high bit error rates. The term high here is relative. For example, when the BI calculation is taken over 80 frames, a bit error rate (BER) greater than 0-3 is high, in that for values of BER above this level there is noticeable isparity between the BI-base an exact probabilities of bit error in the BI wor. When the number of frames is 970 the transition level is at a BER of 0-3, an so on. The reason for this is given by (3) an (4) an epicte in Fig.4. The analysis presente here can be applie in SOET systems employing automatic protection switching (AS). Technical stanars exist that specify requirements (an objectives) for eclaring an clearing alarms. Details of SOET frame structure an automatic protection switching can be foun in telecommunication stanars[-4] an other texts [5-8]. The protection mechanism can be of two types [6], the : an :n protection mechanisms. Fig. (a) epicts the : protection architecture where a protection interface is paire with each working interface. Fig. (b) epicts the other the : n protection architecture, consisting of a single protection facility for several working interfaces. In either case, when a working interface fails, traffic is automatically switche over to the protection interface. The faile working facility is marke with an in both halves of Fig.. V. K. Ouol is with the Department of Electrical an Information Engineering, University of airobi, airobi, Kenya ( ext.837, vkouol@uonbi.ac.ke) Cemal Aril is with the ational Acaemy of Aviation, Baku, Aerbaijan The SOET STS- frame structure is given in Fig.A of the Appenix. For the purposes of this paper, the bytes of interest are the three BI bytes B, B an B3 whose scopes are as follows. The B byte is use to etect parity errors per frame. This is one for the first STS- frame in the STS-n multiplexe frame. It monitors section level bit errors. The B byte is use to monitor line-level bit errors, an the B3 byte is use to monitor path-level bit errors, inclusive of the path overhea. A single interleave parity byte is use to provie error monitoring across a particular segment along the en-to-en SOET path. This parity byte performs a parity check on the previous Synchronous Transport Signal level (STS-) frame. During the parity check, the first bit of the BI octet is a parity check on the first bit of all octets of the previously scramble STS- frame. The secon bit of the BI octet is use exactly the same way, i.e. it is a parity check on the secon bits of each octet of the previous STS- frames, an similarly for the other bits. Hea en switch Signal iverte Hea en switch Signal iverte Tail en switch protection facility (a) : rotection Tail en switch protection facility (a) : n rotection Fig. protection switching When the bit errors excee certain threshols, i.e. when there is an excessive bit error rate conition, the system may eclare an alarm. These conitions when present must be eclare within time limits specifie by telecommunication stanars [,3]. The rest of the paper is organie as follows: Section II presents a iscussion an etermination of the bit error probability in the BI wor. Section III presents the eclaration of the alarm using a sliing winow, where first International Scholarly an Scientific Research & Innovation 5() scholar.waset.org/ /6347

2 Vol:5, o:, 0 International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 the winow containing the alarm is etermine, an then the alarm is locate within the winow. Section IV iscusses the clearing of the alarm. In a manner parallel to the eclaration of the alarm, the winow containing the alarm clear conition is first etermine, then the alarm clear location is foun within the winow. Section IV presents the results an conclusions. Furthermore, the Appenix provies some information about the SOET STS- frame an some pertinent BI count parameters. II. BIT ERROR ROBABILITY I A BI WORD Fig.. shows the bytes to be inclue in the computation of the bit interleave parity (BI) byte. The bit error for the bit in the secon position is registere only if there is a total of an o number of errors in the bits consiere. If there is only one byte to consier, the probability that there is a bit error in the secon bit position is p. Suppose - bytes have been consiere so far Byte Byte - Byte p p Fig. Bit interleave parity probability of bit error for the bit in the secon position of the BI wor The probability that that a bit error is registere for the secon position is obtaine by consiering two isjoint events. Either there is an error at the en of the previous stage (-) an no error in byte- or there is no error at the en of the previous stage an an error at byte-. These two give rise to the equation p ( p) ( ) p () which is quickly rearrange as pp p () with the initial conition that 0 =0. It is easily verifie that the solution to the above equation is p (3) Here is the number of bits inclue in the calculation. The exact bit error probability consiers that there are bits an the probability of bit error is then the complement of the probability of no bit error in any of the secon position bits. Accoringly the exact probability that a bit in the BI wor is registere as being in error is given by p (4) In aition to the bytes inclue in (3) an (4) there is further potential for bit errors in the BI byte of the current frame. This is compare with a calculate version of the BI byte. Thus there are + bytes to inclue in (3) an (4). These equations are compare in Fig.3 for some typical values of liste in Table AI. For =80, the BI-base (3) an the exact (4) values begin to iverge at a bit error rate of 0 4, with the BIbase value settling at 0.5, an the exact one at.0. This tren is seen for the other values of. Inee the point of separation of the two values gets smaller as increases, but for all the cases, the exact traces go to.0 whereas the BIbases trace goes to 0.5. These can be erive from the expressions alreay given by letting p ten to in (3) an (4), respectively. BI Wor Bit Error robability =87480 =970 =87480 =970 =80.0E-06.0E-05.0E-04.0E-03.0E-0 Actual BER =80 BI-Base Fig. 3 Variation of BI wor bit error probability with the number of bits use in the BI calculation. Traces for three values of are shown It is also observe that as increases, the curves in Fig.3 shift towars the left, the BI-base traces approaching a limit of 0.5, an the exact traces approach unity. Despite the above isparity of (3) an (4), it is possible to obtain working bit error rate monitoring methos base on BI calculations using a winow of an appropriate sie an two threshols one for eclaring alarm, an the other for clearing. A frame is consiere errore when it has more than one bit error in the BI byte. The probability that a frame is errore is then 8 8 8m m (5) m m A winow of sie M frames is use an an alarm is eclare if there are or more errore frames in the winow. Conversely, once an alarm is eclare the alarm is cleare when there are or more non-errore frames in a winow of sie M frames. Alarm eclaration an clearing is cyclic International Scholarly an Scientific Research & Innovation 5() scholar.waset.org/ /6347

3 Vol:5, o:, 0 International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 process which consists of four stages as shown in Fig.5. The analysis uses a sliing winow in monitoring the BI bit errors. A sliing winow has many avantages over a jumping winow. It is therefore wiely use in many stanar network protocols, an several authors[9-] have provie recent analysis of the performance protocols base on sliing winows. For the present objective, a sliing winow will have the avantage over a jumping winow in that the sprea of the errors may traverse the bounary of a winow, an a jumping winow will fail to catch some of those error patterns that lie on the winow bounaries. A sliing winow, on the other han will not miss such error patterns. Returning to Fig.4, it is evient that first there is a hunt for the winow containing the alarm, then the alarm is eclare within the winow. For clearing the alarm, there is a hunt for the winow containing the alarm clear, followe by locating the alarm clear conition within the winow. This cyclic process is repeate for as long as the system runs, being restarte only when the bit error estimates ictate that the winow parameters M, an shoul be change. The etaile escriptions of the components of Fig.3 are given in the sections that follow. III. ALARM DECLARATIO The unconitional event that a winow contains an alarm is equivalent to the event that there are at least errore frames in a winow of sie M. This gives the probability Declr of alarm eclaration as M M m M m (6) Declr m m To locate, eclare an clear alarms, the implementation of the bit error rate monitoring uses a sliing winow as epicte in Fig.4. Hunt for winow with alarm - Q Alarm Declare M Hunt for clear winow Locate Alarm M Alarm Cleare Locate Alarm Clear Fig. 4 Alarm eclaration an clearing Q - The process involves first hunting for the winow containing the alarm, followe by locating the alarm within the winow. The first block after the summing point represents the fact that the system waits for the occurrence of an errore frame. The length K D of time in frames up to an incluing the errore frame is a geometrically istribute ranom variable satisfying the istribution k rob K D k k (7) whose moment generating function KD () is (8) KD Once the first errore frame is foun, the system examines the subsequent frames keeping two counts, the number of frames examine, an the accumulate number of errore frames so far. If the number of frames reaches the winow sie before the number of errore frames reaches the threshol, the search begins afresh. This is the event that there are fewer than errore frames in a winow of sie M, where appropriate allowance has been mae for the fact that one errore frame occurs at the beginning of the target winow. Thus the probability Q that the search begins afresh is then given by M M m m Q (9) m0 m If the search results in locating the winow with alarm, then point at which the alarm is eclare has to be etermine. There is no nee to examine the whole winow if the require number of errore frames has been reache. Thus the length of time to the eclaration of the alarm is another ranom variable. A. Waiting Time to Reach Winow Containing Alarm At this point it is reasonable to obtain an expression for the waiting time to the winow containing the alarm. This is one via its moment generating function T WD (). The feeback iagram of Fig.4 can be use to give Q T (0) WD ( ) M Q It is note that for =, the enominator of this expression is ero for Q =. The quantity in (0) being a moment generating function is require to be analytic insie an on the unit isc { : }. This requirement will be violate for Q =. Inee, the mean time to the alarm winow which is obtaine as M Q T () ( ) WD Q reveals that for values of Q close to, the system waiting time on average will be unacceptably large. Thus, the winow parameters M an of the system must be chosen to ensure that Q is far enough away from. The observations mae on (0) an () are a consequence of the sliing winow International Scholarly an Scientific Research & Innovation 5() 0 94 scholar.waset.org/ /6347

4 Vol:5, o:, 0 International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 mechanism; it remains to be seen if the BI calculations are sensitive to this fact. This is eferre to the results presente later. B. Locating the Alarm Within Winow At this point the winow containing the alarm has been foun. An accumulate count is kept within the winow the errore frames, an the alarm is eclare as soon as the number of errore frames reaches the threshol ; there is no nee to reach the en of the winow. This is one in the block containing the moment generating function T D () in Fig.4. The probability Declr (j) of eclaring an alarm after j frames is the event that in the preceing j frames there are errore frames, followe by an errore frame. The probability of this event is then j j ( ) Declr j () To obtain the probability of an alarm eclaration the expression in () can be summe for j = to M, an incorporating a multiplying factor to cater for the fact that () is a conitional probability. This is a much longer metho than the expression given in (6), which consiers that there at least errore frames in the winow for the alarm to be eclare. Given that the winow containing the alarm has been locate, the remaining waiting time T D to alarm eclaration is then given via its moment generating T D () as M j j j TD (3) j The ranom variables T WD an T D referre to in (0) an (3), respectively, are statistically inepenent. The total time T to eclare the alarm is the sum of these two. ( ) T T T (4) WD D Substituting (0) an (3) in (4) gives the mean alarm eclaration time as M Q M (5) T B, j, Q j where is B( -, j, ) is the binomial probability of - successes in j Bernoulli trials, an is the probability of success. This quantity along with the corresponing one for clearing of the alarm can be evaluate for ifferent system parameters. IV. ALARM CLEARIG The unconitional probability that a winow of sie M contains an alarm clear conition is equivalent to the event that there are at least non-errore frames in a winow of sie M. In an analogous manner to (6), this gives the probability of alarm clearing as M M M m m (6) Clear m m The same sliing winow above is use in the bit error rate monitoring to locate the conition to clear the alarm. This section is very similar to the preceing one, the ifference being that whereas the hunt for the alarm counts the errore frames, here it is the non-errore frames that are counte, an the parameter is now replace by. With reference to Fig.4, the parameter is replace by, an Q is replace by, efine similar to (9) as M M m m (7) m0 m an is the probability that there are fewer than - non-errore frames in the winow of sie M-. As before the length K C of time in frames up to an incluing the terminating non-errore frame is a geometrically istribute ranom variable satisfying the istribution k rob K C k k (8) whose moment generating KC () is (9) KC with the corresponing one for clearing of the alarm, is evaluate for ifferent system parameters. A. Waiting Time to Reach Winow To Clear Alarm Exploiting the similarity with the preceing evelopment, the waiting time T WC in frames require to reach the winow is efine via its moment generating function T WC () as T (8) WC M The mean time to the clear winow is obtaine as M T (9) ( ) WC As before, it also hols here that for values of close to, the system waiting time on average will be unacceptably large, which unerscores once more the fact that the winow parameters M an must be chosen to ensure that is far enough away from. B. Locating the Alarm Clear Within Winow The winow containing the alarm clear conition having been locate, the alarm clear is inicate as soon as the number of non-errore frames reaches the threshol ; there is no nee to reach the en of the winow. The probability Clear (j) of clearing an alarm after j frames is the event that in the preceing j frames there are non-errore frames, followe by a non-errore frame. The probability of this event is then j j ( ) Clear j (0) The probability that an alarm is cleare within the winow can be obtaine by summing the above for j = to M. That is M () ( j) Clear Clear j The expressions in (6) an () are equivalent since they International Scholarly an Scientific Research & Innovation 5() 0 94 scholar.waset.org/ /6347

5 Vol:5, o:, 0 International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 refer to the same events. The results provie here are obtaine base on (6). In practice the parameters M, an are chosen to ensure that the eclaration an clearing probabilities meet certain requirements establishe by applicable telecommunication stanars such as [] an [3]. Having locate the winow containing the alarm clear, the next task is to locate the alarm clear within the winow. The remaining waiting time T C to alarm clearing is then given via its moment generating T C () as M j j j TC () j The total time to clear the alarm T is the sum of these two T T T (3) WC C Substituting (8) an () in (3) gives the mean time in frames to alarm clearing as M M T j B, j, (4) where B(, j, ) is the binomial probability of successes in j trials, an is the probability of success. This quantity along with the corresponing one for the eclaration of the alarm can be evaluate for ifferent system parameters an the results compare for both the BI-base an exact methos. Rather than clutter the presentation with too many results, only those corresponing to =80 are given here. V. RESULTS AD COCLUSIO From Fig.4 it has alreay been observe that the two expressions (3) an (4) for the bit error probability in the BI byte eviate as the prevailing bit error rate increases, giving the first inication that the BI-base results may iffer from the actual situation. When the bit error rate is low, the eviation is small. The onset of eviation epens on, an ecreases with, the number of frames use in the BI calculation. The sliing winow parameters = 80, M = 64, = 49, an = 3 were use to generate the results presente here. Table I an Fig.5 give the results for the eclaration times in secons for sliing winow parameters inicate. TABLE I ALARM DECLARATIO TIMES FOR = 80 BER Declaration Times [s] BI-Base.00E-04.00E+7.60E+3.78E E E+ 3.6E E E E E E-03.00E E E-03.78E E E E E E E E E-03.00E E E-03 The eclaration times while compliant with the requirements of the telecommunication stanars, show a eviation. For example when BER = the eclaration times are 4.50 ms (BI) an 6.7 ms (exact). This inicates that the BI-base system eclares the alarm later than shoul be case. This poses a challenge to the stanars evelopers to ensure that the alarm eclaration threshols are set so that even though the alarm are set later the results can still be use to guarantee acceptable network performance. Declaration Time [s] BI-Base E-04.0E-03.0E-0.0E-0 Actual BER Fig. 5 Alarm eclaration tims for = 80 Table II an Fig.6 give the alarm clearing times. Evient from theses results is the fact that the clearing times higher values of the bit error rate are much lower for the BI-base sliing winow that those obtaine by the exact calculations. Inee for BER = the clearing times are secons (BI) an secons (exact). This inicates that the BI-base system clears the alarm too soon. Whereas the exact calculation inicates a clearing time of over hours, the BI-base version clears the alarm in just over 3 minutes. TABLE II ALARM CLEARIG TIMES FOR = 80 BER Clearing Times [s] BI-Base.00E E E-03.78E E E E E E E E E-03.00E E E-0.78E E E E E E E E E+4.00E E E+5 The scenario presente here is that the BI-base result inicates an alarm eclaration later an clears the alarm too soon. The sliing winow algorithm to be establishe to take into account the fact that there may be elaye eclaration an false clearing of the alarms. Fortunately, the stanars evelopers have built some hysteresis into the requirements [] to guar against false clearing. The benefits of the analysis are to the stanars evelopers International Scholarly an Scientific Research & Innovation 5() scholar.waset.org/ /6347

6 Vol:5, o:, 0 International Science Inex, Electrical an Computer Engineering Vol:5, o:, 0 waset.org/ublication/6347 who must set the requirements to ensure that when BI-base error monitoring is employe there will be isparities with the exact results. For the implementers of the network elements, the analysis presente here will be useful in setting the sliing winow system parameters, to ensure that acceptable performance is achieve. Clearing Time [s].00e+05.00e+04.00e+03.00e+0.00e+0.00e+00.00e-0.00e-0.00e-03.00e-04.00e-04.00e-03.00e-0 Actual BER BI-Base Fig. 6 Alarm clearing times for =80 AEDIX Fig.A shows the SOET STS- frame inicating the overhea bytes an their functions. Table AI shows the values of the number of frames use in the BI calculations. The ata in this table can be use together with (3) an (4) to obtain the probability that a bit in the BI wor is inicate as being in error. 9 rows 3 rows Transport Overhea 3 bytes Section Overhea Line Overhea Synchronous payloa envelope (SE) 87 bytes A Framing A Framing C STS- ID J Trace B BI-8 E Orerwire F User B3 BI-8 D Datacom D Datacom D3 Datacom ath Overhea C Signal Label H ointer H ointer H3 ointer Action G ath Status B BI-8 K AS K AS F User D4 Datacom D5 Datacom D6 Datacom D7 Datacom D8 Datacom D9 Datacom D0 Datacom D Datacom D Datacom H4 Multiframe Z3 Growth Z4 Growth A - Framing A - Framing E Orerwire Z5 Growth Fig. A SOET STS- Overhea bytes an their functions TABLE AI. BI COUT ARAMETERS BI Type Signal B STM-0 / STS- 80 B STM- / STS-3,430 B STM-4 / STS- 9,70 B STM- / STS-48 9,60 B STM-48 / STS-9 87,480 B All signals 80 B3 VC-3 / STS- 783 B3 VC-4 / STS-3c,349 RERECES [] SOET Transport Systems; Common Generic Criteria, Bellcore publication GR-53 CORE, Section 5.3 [] ASI T.05: SOET - Basic Description incluing Multiplex Structure, Rates an Formats [3] ASI T.05.0: SOET - Automatic rotection Switching [4] ASI T.05.07: SOET - Sub-STS- Interface Rates an Formats Specification [5] International Engineering Consortium, SOET Tutorial, Available [6] U. Black, S. Waters, Sonet an T: Architectures for Digital Transport etworks, n e, rentice Hall, 005 [7] M.. Ellanti, S.S. Gorshe, L.G. Raman, W.D. Grover, ext Generation Transport etworks: Data,Mmanagement, an Control lanes, Springer, 005 [8] H. G. erros, Connection-Oriente etworks: SOET/SDH, ATM, MLS, an Optical etworks, Wiley 005 [9] D. Chkliaev, J. Hooman,E. e Vink, Formal Verification of an Improve Sliing Winow rotocol in roc. 3r rogress Symposium on Embee Systems, 00, Utrecht, The ethelans, October 00, pages [0] Mark A. Smith an ils Klarlun. Verification of a Sliing Winow rotocol Using IOA an MOA. FORTE/STV 000, isa, Italy, October 000, pages [] Eric Maelaine an Diier Vergamini. Specification an Verification of a Sliing Winow rotocol in LOTOS. FORTE '9, Syney, Australia, ovember 99, pages [] Y. Deng, Z. Huang, Moeling an erformance Analysis of a Sliing Winow rotocol roc. 4th rogress Symposium on Embee Systems, pp , Utrecht, STW, The etherlans October 003 Vitalice K. Ouol receive his pre-university eucation at Alliance High School in Kenya. In 98 he was aware a CIDA scholarship to stuy electrical engineering at McGill University, Canaa, where he receive the B.Eng. (Hons.) an M.Eng. egrees in 985 an 987, respectively, both in electrical engineering. In June 99, he receive the h.d. egree in electrical engineering at McGill University. He was a research associate an teaching assistant while a grauate stuent at McGill University. He joine MB Technologies, Inc. in 989, where he participate in a variety of projects, incluing meteor burst communication systems, satellite on-boar processing, low probability of intercept raio, among others. In 994 he joine ITELSAT where he initiate research an evelopment work on the integration of terrestrial wireless an satellite systems. After working at COMSAT Labs. ( ) on VSAT networks, an TranSwitch Corp.(998-00) on prouct efinition an architecture, he returne to Kenya, where since 003 he has been with Department of Electrical an Information Engineering, University of airobi. Dr. Ouol was a two-time recipient of the Douglas tutorial scholarship at McGill University. He is currently chairman, Department of Electrical an Information Engineering, University of airobi. His research interests inclue performance analysis, moeling an simulation of telecommunication systems, aaptive error control, feeback communication. Cemal Aril is with the ational Acaemy of Aviation, Baku, Aerbaijan International Scholarly an Scientific Research & Innovation 5() scholar.waset.org/ /6347