We consider multirate loss models of quasi‐random arriving calls with elastic bandwidth requirements and fixed bandwidth allocation during service. Calls may retry several times upon arrival (requiring less bandwidth each time) in order to be accepted for service. Alternatively, new calls may request less bandwidth according to the occupied link bandwidth indicated by threshold(s).
In the finite single‐retry model (f‐SRM), a single link of capacity b.u. accommodates calls of service‐classes under the CS policy. Calls of service class come from a finite source population . The mean arrival rate of service‐class idle sources is given by , where is the number of in‐service calls and is the arrival rate per idle source. The offered traffic‐load per idle source of service‐class is given by (in erl). Note that if for , and the total offered traffic‐load remains constant, then the call arrival process is Poisson. A new call of service‐class has a peak‐bandwidth requirement of b.u. and an exponentially distributed service time with mean . If the initially required b.u. are not available in the link, the call is blocked and immediately retries to be connected in the system with b.u. while the mean of the new service time increases to so that the product bandwidth requirement by service time remains constant [1]. If the b.u. are not available the call is blocked and lost. The CAC mechanism of a call of service‐class is identical to that of Figure 2.2 of the SRM, i.e., a new call of service‐class is blocked with b.u. if and is accepted with if , where and are the in‐service calls of service‐class accepted with b.u., respectively.
The comparison of the f‐SRM with the EnMLM reveals similar differences to those described in Section 2.1.1 for the SRM and the EMLM.
To describe the analytical model in the steady state, let us concentrate on a single link of capacity C b.u. that accommodates only two service‐classes with the following traffic characteristics: . Blocked calls of service‐class 2 may retry with parameters while blocked calls of service‐class 1 do not retry. Although the f‐SRM does not have a PFS, we assume that the LB equation (6.33), proposed in the EnMLM, does hold, that is [ 1]:
or, due to the fact that in the equivalent stochastic system:
This assumption is important for the derivation of an approximate but recursive formula for the . If , when a new call of service‐class 2 arrives in the system this call is blocked and retries to be connected with b.u. If , the retry call will be accepted in the system. To describe this situation we need an additional LB equation [ 1]:
where is the offered traffic‐load per idle source of service‐class 2 with , and is the mean number of service‐class 2 calls accepted in state with .
Multiplying (7.3) with , we have for :
Multiplying both sides of (7.2) with , we have for :
Adding (7.4) to (7.5), and since , we obtain:
Apart from the assumption of the LB equation ( 7.3), another approximation is necessary for the recursive calculation of :
In (7.7), the value of is considered negligible compared to when . This is the migration approximation (see Section 2.1.2.1). Due to ( 7.7), equation ( 7.5) is written as (for :
The combination of (7.6) and (7.8) gives an approximate formula for the determination of in the f‐SRM, assuming that only calls of service‐class 2 can retry [ 1]:
where , and when (otherwise ).
The generalization of (7.9) in the case of service‐classes, where all service‐classes may retry, is as follows [ 1]:
where when (otherwise ).
Note that if , for , and the total offered traffic‐load remains constant, then we have the recursion (2.10) of the SRM.
The following performance measures can be determined based on (7.10):
The determination of via ( 7.10), and consequently of all performance measures, requires the values of and , which are unknown. In [ 1–3] give a method for the determination of and in each state via an equivalent stochastic system, with the same traffic parameters and the same set of states, as already described for the proof of (6.27) in the EnMLM. However, the state space determination of the equivalent system is complex, even for small capacity systems that serve many service‐classes with the ability to retry.
Consider a link of capacity b.u. and two service‐classes whose calls require and b.u., respectively. The offered traffic load per idle source is erl, while the number of sources is . Blocked calls of service‐class 1 retry with reduced bandwidth, b.u., and increased offered traffic erl, so that the total offered traffic load per idle source remains the same . Due to the migration approximation, retry calls of service‐class 1 are assumed to be negligible when the occupied link bandwidth . Taking into account the migration approximation assumption, the state space consists of eight states , presented in Table 7.1 together with the respective occupied link bandwidth and the blocking states . According to Table 7.1, the values of appear more than once, and therefore it is impossible to use ( 7.10) directly for the calculation of , e.g., the value corresponds to both and . To overcome this, an equivalent stochastic system should be determined with the following three characteristics:
In this example, an approximate solution to the initial system is , and ; for these values we present in the last column of Table 7.1 the unique values of the equivalent occupied link bandwidth, . The resultant TC probabilities are , and . Note that every system that is a multiple of , and (e.g., , and ) gives exactly the same TC probabilities.
Table 7.1 State space, , , and blocking states (Example 7.2).
0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 3 | 3000 | * | |||||
0 | 1 | 0 | 2 | 2000 | 1 | 0 | 1 | 4 | 4001 | * | * | ||||
0 | 2 | 0 | 4 | 4000 | * | * | 1 | 0 | 2 | 5 | 5002 | * | * | * | |
0 | 2 | 1 | 5 | 5001 | * | * | * | 1 | 1 | 0 | 5 | 5000 | * | * | * |
As already mentioned, the determination of an equivalent system is complex, especially when calls retry many times. To circumvent this problem we present, in the next section, the algorithm of Section 6.2.2.3 (initially proposed for the EnMLM) for the f‐SRM.
Contrary to ( 7.10), which requires enumeration and processing of the state space, the algorithm presented herein is much simpler and easy to implement [4,5].
Consider again Example 7.1 . Apply the algorithm of Section 7.1.2.3 for the determination of TC probabilities.
Based on (2.10), we have the following normalized values of :
(compare with the exact 0.190881) |
(compare with the exact 0.540931) |
(compare with the exact 0.364034) |
In the f‐SRM under the BR policy (f‐SRM/BR), b.u. are reserved to benefit calls of all other service‐classes apart from service‐class k. The application of the BR policy in the f‐SRM is similar to that of the SRM/BR as the following example shows.
Consider again Example 7.1 and let the BR parameters b.u. and b.u. so that . In that case, equalization of TC probabilities is achieved between new calls of service‐class 1 and new (not retry) calls of service‐class 2.
The solution of this linear system is:
Then, based on the values of , we obtain the values of :
(0.548536 in the SRM/BR, see Example 2.3) |
The of service‐class 2 calls, , when they require b.u., is:
(0.548536 in the SRM/BR) |
The TC probabilities of service‐class 2 calls, , when they require b.u., are:
(0.301968 in the SRM/BR) |
In the f‐SRM/BR, the unnormalized values of can be calculated in an approximate way according to the Roberts method (see Section 1.3.2.2). Based on that method, we can either find an equivalent stochastic system (which is complex) [ 3] or apply an algorithm similar to the one presented in Section 7.1.2.3 . Due to its simplicity, the latter is adopted herein with the following modifications.
The following performance measures can be determined based on (7.17):
Consider again Example 7.4 . Apply the algorithm of Section 7.2.2.1 for the determination of TC probabilities.
Based on (2.18), we have the following normalized values of :
The approximate TC probabilities are:
(compare with the exact 0.539377) |
(compare with the exact 0.539377) |
(compare with the exact 0.285796) |
In the finite multi‐retry model (f‐MRM), calls of service‐class k can retry more than once to be connected in the system [2– 5]. Let be the number of retrials for calls of service‐class k and assume that , where is the required bandwidth of a service‐class k call in the th retry, . Then a service‐class k call is accepted in the system with b.u. if .
Consider a link of b.u. The link accommodates quasi‐random arriving calls of two service‐classes. Calls of service‐class 1 require b.u. while calls of service‐class 2 require . Blocked calls of service‐class 2 can retry two times with b.u. and b.u. Let the number of sources be , the arrival rate per idle source be and sec, sec, and sec (in this example we do not assume that ).
The solution of this linear system is:
Based on the values of , we have:
(0.545687 in the MRM, see Example 2.6) |
The TC probability of the first retry of service‐class 2 is given by:
(0.876443 in the MRM) |
The TC probability of the second retry of service‐class 2, with b.u., is given by:
(0.725882 in the MRM) |
The TC probability of service‐class 2, with b.u., is given by:
(0.545687 in the MRM) |
The link utilization is determined by:
b.u. (2.148 in the MRM) |
(0.62262 in the MRM). |
Following the analysis of Section 7.1.2.1, in the f‐MRM not only LB is assumed but also the migration approximation, that is, the mean number of service‐class k calls in state , accepted with b.u., is negligible when . This means that service‐class k calls with are limited in the area . The are determined by [ 2]:
where , and when (otherwise ).
Note that if for , and the total offered traffic‐load remains constant, then we have the recursion (2.22) of the MRM [6]. In addition, if calls may retry only once, then the SRM results [ 6].
The following performance measures can be determined based on (7.21):
The determination of via ( 7.21), and consequently of all performance measures, requires the values of and , which are unknown. In [ 2], there is a method for the determination of and in each state via an equivalent stochastic system, with the same traffic parameters and the same set of states. However, since this method is complex, we adopt the algorithm of Section 7.3.2.3 (see below), which is similar to the algorithm of Section 7.1.2.3 of the f‐SRM.
The algorithm for the calculation of in the f‐MRM can be described as follows [ 4, 5]:
Consider again Example 7.6 (, ). Apply the algorithm of Section 7.3.2.3 for the determination of TC probabilities, including the retry probabilities and the link utilization.
Based on (2.22), we have the following normalized values of :
(compare with the exact 0.5274) |
(compare with the exact 0.867574) |
(compare with the exact 0.709409) |
(compare with the exact 0.5274) |
The link utilization is determined by:
b.u. (compare with the exact 2.104) |
Compared to the f‐SRM/BR, in the f‐MRM under the BR policy (f‐MRM/BR), blocked calls of service‐class k can retry more than once to be connected in the system.
Consider again Example 7.6 (, ) and let the BR parameters b.u. and b.u. so that .
The solution of this linear system is:
Based on the values of , we have:
(compare with the exact 0.819672 in the MRM/BR, see Example 2.8) |
The TC probabilities of service‐class 2, with , are given by:
(compare with the exact 0.819672 in the MRM/BR) |
The TC probabilities of service‐class 2, with b.u., are given by:
(compare with the exact 0.606557 in the MRM/BR) |
The TC probabilities of service‐class 2, with b.u., are given by:
(compare with the exact 0.514754 in the MRM/BR) |
The link utilization is determined by:
b.u. (compare with the exact 1.941 b.u. in the MRM/BR) |
Based on the Roberts method and the algorithm of Section 7.2.2.1 of the f‐SRM/BR, the unnormalized values of the link occupancy distribution, , in the f‐MRM/BR can be determined by the following algorithm:
The following performance measures can be determined based on (7.28):
Consider again Example 7.8 (, ). Apply the algorithm of Section 7.4.2.1 for the determination of TC and retry probabilities.
The approximate TC probabilities and retry probabilities are:
(compare with the exact 0.814167) |
(compare with the exact 0.814167) |
(compare with the exact 0.597726) |
(compare with the exact 0.50153) |
Consider a link of capacity b.u. and service‐classes with and b.u. The offered traffic‐loads are and erl. Blocked calls of service‐class 1 reduce their bandwidth requirement twice, from 10 to 8 and finally to 6 b.u. The corresponding retry offered traffic‐loads are and erl. Blocked calls of service‐class 2 reduce their bandwidth once, from 7 to 4 b.u. The retry offered traffic‐load of service‐class 2 is erl. The number of sources for both service‐classes is . Based on the above, we have a f‐MRM system. The equivalent stochastic system used for the TC probabilities calculation is , and . Table 7.2 shows the state space, which consists of 30 states.
Table 7.2 The state space, , and the blocking states (Example 7.10).
0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
0 | 1 | 0 | 0 | 0 | 7 | 7001 | ||
0 | 2 | 0 | 0 | 0 | 14 | 14002 | ||
0 | 3 | 0 | 0 | 0 | 21 | 21003 | ||
0 | 4 | 0 | 0 | 0 | 28 | 28004 | ||
0 | 5 | 0 | 0 | 0 | 35 | 35005 | ||
0 | 6 | 0 | 0 | 0 | 42 | 42006 | ||
0 | 6 | 1 | 0 | 0 | 50 | 50006 | * | * |
0 | 7 | 0 | 0 | 0 | 49 | 49007 | * | * |
1 | 0 | 0 | 0 | 0 | 10 | 10000 | ||
1 | 1 | 0 | 0 | 0 | 17 | 17001 | ||
1 | 2 | 0 | 0 | 0 | 24 | 24002 | ||
1 | 3 | 0 | 0 | 0 | 31 | 31003 | ||
1 | 4 | 0 | 0 | 0 | 38 | 38004 | ||
1 | 5 | 0 | 0 | 0 | 45 | 45005 | * | |
1 | 5 | 0 | 0 | 1 | 49 | 49005 | * | * |
2 | 0 | 0 | 0 | 0 | 20 | 20000 | ||
2 | 1 | 0 | 0 | 0 | 27 | 27001 | ||
2 | 2 | 0 | 0 | 0 | 34 | 34002 | ||
2 | 3 | 0 | 0 | 0 | 41 | 41003 | ||
2 | 3 | 1 | 0 | 0 | 49 | 49003 | * | * |
2 | 4 | 0 | 0 | 0 | 48 | 48004 | * | * |
3 | 0 | 0 | 0 | 0 | 30 | 30000 | ||
3 | 1 | 0 | 0 | 0 | 37 | 37001 | ||
3 | 2 | 0 | 0 | 0 | 44 | 44002 | ||
3 | 2 | 0 | 0 | 1 | 48 | 48002 | * | * |
3 | 2 | 0 | 1 | 0 | 50 | 50002 | * | * |
4 | 0 | 0 | 0 | 0 | 40 | 40000 | ||
4 | 1 | 0 | 0 | 0 | 47 | 47001 | * | * |
5 | 0 | 0 | 0 | 0 | 50 | 50000 | * | * |
The corresponding MRM system used for the calculation of and and consequently of TC probabilities (according to Section 7.3.2.3 ) is . In Table 7.3, we present the analytical TC probabilities obtained from the equivalent stochastic system and the algorithm of Section 7.3.2.3 together with the corresponding simulation results. The latter are mean values of seven runs with 95% confidence interval. At each point (P) of Table 7.3 the values of , and are constant while those of and are increased by and , respectively (for and erl). Based on Table 7.3, one observes that the algorithm of Section 7.3.2.3 gives almost the same results as the equivalent stochastic system method and satisfactory results compared to simulation.
Table 7.3 Analytical and simulation results of TC probabilities (Example 7.10).
Equivalent stochastic system | Algorithm (Section 7.3.2.3 ) | Simulation | ||||
P | (%) | (%) | (%) | (%) | (%) | (%) |
1 | 3.05 | 2.03 | 3.05 | 2.03 | 3.23 0.49 | 1.83 0.13 |
2 | 4.32 | 2.72 | 4.32 | 2.72 | 4.02 0.32 | 2.44 0.12 |
3 | 5.78 | 3.50 | 5.78 | 3.50 | 5.19 0.32 | 3.20 0.24 |
4 | 7.39 | 4.36 | 7.39 | 4.36 | 7.01 0.30 | 3.76 0.29 |
5 | 9.12 | 5.30 | 9.13 | 5.30 | 8.16 0.49 | 4.71 0.35 |
6 | 10.94 | 6.31 | 10.94 | 6.31 | 9.58 0.44 | 5.67 0.22 |
In the finite single‐threshold model (f‐STM), the requested b.u. and the corresponding service time of a new call are related to the occupied link bandwidth and a threshold . Specifically, the following CAC is applied. When , then a new call of service‐class k is accepted in the system with its initial requirements . Otherwise, if , the call tries to be connected in the system with , where and , so that the product bandwidth requirement by service time remains constant. If the b.u. are not available the call is blocked and lost. The call arrival process is the quasi‐random, i.e., calls of service class come from a finite source population . By denoting as the arrival rate per idle source of service‐class k, the offered traffic‐load per idle source is given by (in erl).
The comparison of the f‐STM with the f‐SRM reveals similar differences to those described in Section 2.5.1 for the STM and the SRM.
Consider again Example 7.1 () and let b.u.
The solution of this linear system is:
Then, based on the values of , we obtain the values of :
(0.165955 in the STM, see Example 2.11) |
The TC probabilities of service‐class 2 calls, , when they require b.u., are:
(0.382997 in the STM) |
The link utilization is determined by:
b.u. (2.685 in the STM) |
To derive a recursive formula for the calculation of , we concentrate on a single link of capacity b.u. that accommodates two service‐classes with the following traffic characteristics: . If upon the arrival of a service‐class 2 call, then this call requests b.u. and . No such option is considered for calls of service‐class 1.
Although the f‐STM does not have a PFS, we assume that the LB equation (6.33) does hold for calls of service‐class 1, for :
or, due to the fact that in the equivalent stochastic system, :
For calls of service‐class 2, we assume the existence of LB between adjacent states:
where is the offered traffic‐load per idle source of service‐class 2 with and is the mean number of service‐class 2 calls accepted in the system with in state .
Equations (7.32)–(7.34) lead to the following system of equations:
For (7.35), we adopt the (migration) approximation that the value of in state is negligible when . For (7.37), we adopt the (upward migration) approximation that the value of in state is negligible when . Based on these approximations, we have (for ):
where , and .
In the general case of service‐classes, the formula for is the following [ 2]:
where ,
Note that if for , and the total offered traffic‐load remains constant, then we have the recursion (2.38) of the STM.
The following performance measures can be determined based on (7.39):
The determination of via ( 7.39), and consequently of all performance measures, requires the value of and , which are unknown. In 2, 3, there is a method for the determination of and in state via an equivalent stochastic system. However, the state space determination of the equivalent system can be complex (as in the case of the EnMLM and the f‐SRM) even for small capacity systems that serve many service‐classes.
Consider Example 7.2 () and let . Due to the migration approximation, calls of service‐class 1 with are assumed to be negligible when . In addition, due to the upward migration approximation, calls of service‐class 1 with are assumed to be negligible when . Taking into account these approximations, the state space consists of 11 states , presented in Table 7.4 together with the respective occupied link bandwidth and the blocking states . According to Table 7.4, the values of appear more than once, and therefore it is impossible to use directly ( 7.39) for the calculation of . To overcome this, an equivalent stochastic system should be determined with the following three characteristics:
In this example, the values , and are an approximate solution to the initial system; for these values we present in the last column of Table 7.4 the unique values of the equivalent occupied link bandwidth, . The resultant TC probabilities are % and %.
Table 7.4 The state space, , and the blocking states (Example 7.12).
0 | 0 | 0 | 0 | 0 | ||
0 | 1 | 0 | 2 | 2000 | ||
0 | 1 | 1 | 3 | 3001 | ||
0 | 1 | 2 | 4 | 4002 | * | |
0 | 1 | 3 | 5 | 5003 | * | * |
0 | 2 | 0 | 4 | 4000 | * | |
0 | 2 | 1 | 5 | 5001 | * | * |
1 | 0 | 0 | 3 | 3000 | ||
1 | 0 | 1 | 4 | 4001 | * | |
1 | 0 | 2 | 5 | 5002 | * | * |
1 | 1 | 0 | 5 | 5000 | * | * |
To circumvent the determination of an equivalent system, in the next section we modify the algorithm of Section 6.2.2.3 (initially proposed for the EnMLM) to fit to the f‐STM.
Contrary to ( 7.39), which requires enumeration and processing of the state space, the algorithm presented herein is much simpler and easy to implement [ 5].
Consider again Example 7.11 (). Apply the algorithm of Section 7.5.2.3 for the determination of TC probabilities.
(compare with the exact 0.152594) |
(compare with the exact 0.365804) |
In the f‐STM under the BR policy (f‐STM/BR), b.u. are reserved to benefit calls of all other service‐classes apart from service‐class k. The application of the BR policy in the f‐STM is similar to that of the STM/BR as the following example shows.
Consider again Example 7.11 () and let the BR parameters be b.u. and b.u. so that .
The solution of this linear system is:
Then, based on the values of , we obtain the values of :
(0.545538 in the STM/BR, see Example 2.13) |
The TC probabilities of service‐class 2 calls, , when they require b.u., are:
(0.313991 in the STM/BR) |
The link utilization is determined by:
b.u. (2.512 in the STM/BR) |
In the f‐STM/BR, the unnormalized values of can be calculated in an approximate way according to the Roberts method. Based on that method, we apply an algorithm similar to the one presented in Section 7.5.2.3 , which is modified as follows:
The following performance measures can be determined based on (7.48):
Consider again Example 7.14 (). Apply the algorithm of Section 7.6.2.1 for the determination of TC probabilities.
Based on (2.44), we have the following normalized values of :
The approximate TC probabilities are:
(compare with the exact 0.535982) |
(compare with the exact 0.295331) |
In the finite multi‐threshold model (f‐MTM), there exist different thresholds which are common to all service‐classes. A call of service‐class k with initial requirements can use, depending on the occupied link bandwidth , one of the requirements , where the pair is used when , while ). The maximum possible threshold is , while . As far as the bandwidth requirements of a service‐class k call are concerned, we assume that they decrease as increases, i.e., , while by definition .
To derive a recursive formula for the calculation of , the following LB equations are considered:
and
where is the number of in‐service calls of service‐class accepted in the system with b.u., and .
Similar to the analysis of the f‐STM and based on (7.52) and (7.53), we have the following recursive formula for the calculation of [ 5]:
where , and .
Note that if for , and the total offered traffic‐load remains constant, then we have the recursion (2.51) of the MTM.
The following performance measures can be determined based on (7.54):
The determination of via ( 7.54), and consequently of all performance measures, requires the values of and , which are unknown. To avoid the determination of an equivalent stochastic system we adopt the algorithm of Section 7.7.2.3 (see below), which is similar to the algorithm of Section 7.5.2.3 of the f‐STM.
The algorithm can be described by the following steps [ 5].
In the f‐MTM under the BR policy (f‐MTM/BR), b.u. are reserved to benefit calls of all other service‐classes apart from service‐class k. The application of the BR policy in the f‐MTM/BR is similar to that of the f‐MRM/BR.
Based on the Roberts method and the algorithm of Section 7.6.2.1 of the f‐STM/BR, the unnormalized values of the link occupancy distribution, , in the f‐MTM/BR can be determined as follows:
Based on (7.61) and (2.58), we can determine the following performance measures:
In the finite CDTM (f‐CDTM), the difference compared to the f‐MTM is that different service‐classes may have different sets of thresholds. Each arriving call of a service‐class k may have bandwidth and service‐time requirements, that is, one initial requirement with values and more requirements with values , where , and . The pair is used when , where and are two successive thresholds of service‐class k, while ; the highest possible threshold (other than ) is . By convention, and , while the pair is used when .
Similar to the analysis of the f‐MTM and based on ( 7.54), we have the following recursive formula for the calculation of [ 5]:
where , , and .
Note that if for , and the total offered traffic‐load remains constant, then we have the recursion (2.65) of the CDTM.
The following performance measures can be determined based on (7.64):
The calculation of via ( 7.64), and consequently of all performance measures, requires the determination of an equivalent stochastic system. To avoid it, we adopt the algorithm of Section 7.9.2.3 (see below), which is similar to the algorithm of Section 7.7.2.3 of the f‐MTM.
The algorithm can be described by the following steps [ 5]:
Consider again Example 7.10 (). The values of and are used when and , respectively. Similarly, the value of is used when . The equivalent stochastic system used for the TC calculation in this f‐CDTM example is . We consider three sets of : (i) , (ii) , and (iii) . Table 7.5 shows the various sets of and the corresponding offered traffic‐loads, while Table 7.6 shows the state space, which consists of 53 states. The corresponding CDTM system used in the algorithm of Section 7.9.2.3 is . For the three different sets of , we present in Tables 7.7–7.9 the analytical TC probabilities obtained via the equivalent stochastic system and the algorithm of Section 7.9.2.3 together with the corresponding simulation results. At each point (P) of Tables 7.7– 7.9, we assume that the values of , and are constant while those of and are increased by 0.4/ and 0.7/, respectively, i.e., point 1 is , while point 6 is , .
According to Tables 7.7– 7.9, the analytical TC probabilities obtained via the algorithm of Section 7.9.2.3 are closer to the simulation results (mean values of 7 runs with 95% confidence interval) than the corresponding results obtained by the equivalent stochastic system.
Table 7.5 Sets of sources and offered traffic‐loads per idle source (Example 7.16).
Number of sources | (erl) | (erl) | (erl) | (erl) | (erl) |
Table 7.6 The state space, , and the blocking states (Example 7.16).
0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
0 | 1 | 0 | 0 | 0 | 7 | 7001 | ||
0 | 2 | 0 | 0 | 0 | 14 | 14002 | ||
0 | 3 | 0 | 0 | 0 | 21 | 21003 | ||
0 | 4 | 0 | 0 | 0 | 28 | 28004 | ||
0 | 5 | 0 | 0 | 0 | 35 | 35005 | ||
0 | 5 | 1 | 0 | 0 | 43 | 43005 | ||
0 | 5 | 1 | 0 | 1 | 47 | 47010 | * | * |
0 | 5 | 1 | 1 | 0 | 49 | 49005 | * | * |
0 | 6 | 0 | 0 | 0 | 42 | 42006 | ||
0 | 6 | 0 | 0 | 1 | 46 | 46011 | * | |
0 | 6 | 0 | 0 | 2 | 50 | 50016 | * | * |
0 | 6 | 0 | 1 | 0 | 48 | 48006 | * | * |
1 | 0 | 0 | 0 | 0 | 10 | 10000 | ||
1 | 1 | 0 | 0 | 0 | 17 | 17001 | ||
1 | 2 | 0 | 0 | 0 | 24 | 24002 | ||
1 | 3 | 0 | 0 | 0 | 31 | 31003 | ||
1 | 3 | 1 | 0 | 0 | 39 | 39003 | ||
1 | 3 | 1 | 0 | 1 | 43 | 43008 | ||
1 | 3 | 1 | 0 | 2 | 47 | 47013 | * | * |
1 | 3 | 1 | 1 | 0 | 45 | 45003 | * | |
1 | 3 | 1 | 1 | 1 | 49 | 49008 | * | * |
1 | 4 | 0 | 0 | 0 | 38 | 38004 | ||
1 | 4 | 0 | 0 | 1 | 42 | 42009 | ||
1 | 4 | 0 | 0 | 2 | 46 | 46014 | * | |
1 | 4 | 0 | 0 | 3 | 50 | 50019 | * | * |
1 | 4 | 0 | 1 | 0 | 44 | 44004 | ||
1 | 4 | 0 | 1 | 1 | 48 | 48009 | * | * |
1 | 4 | 0 | 2 | 0 | 50 | 50004 | * | * |
2 | 0 | 0 | 0 | 0 | 20 | 20000 | ||
2 | 1 | 0 | 0 | 0 | 27 | 27001 | ||
2 | 2 | 0 | 0 | 0 | 34 | 34002 | ||
2 | 2 | 1 | 0 | 0 | 42 | 42002 | ||
2 | 2 | 1 | 0 | 1 | 46 | 46007 | * | |
2 | 2 | 1 | 0 | 2 | 50 | 50012 | * | * |
2 | 2 | 1 | 1 | 0 | 48 | 48002 | * | * |
2 | 3 | 0 | 0 | 0 | 41 | 41003 | ||
2 | 3 | 0 | 0 | 1 | 45 | 45008 | * | |
2 | 3 | 0 | 0 | 2 | 49 | 49013 | * | * |
2 | 3 | 0 | 1 | 0 | 47 | 47003 | * | * |
3 | 0 | 0 | 0 | 0 | 30 | 30000 | ||
3 | 1 | 0 | 0 | 0 | 37 | 37001 | ||
3 | 1 | 0 | 1 | 0 | 43 | 43001 | ||
3 | 1 | 0 | 2 | 0 | 49 | 49001 | * | * |
3 | 1 | 0 | 1 | 1 | 47 | 47006 | * | * |
3 | 2 | 0 | 0 | 0 | 44 | 44002 | ||
3 | 2 | 0 | 0 | 1 | 48 | 48007 | * | * |
3 | 2 | 0 | 1 | 0 | 50 | 50002 | * | * |
4 | 0 | 0 | 0 | 0 | 40 | 40000 | ||
4 | 0 | 0 | 0 | 1 | 44 | 44005 | ||
4 | 0 | 0 | 0 | 2 | 48 | 48010 | * | * |
4 | 0 | 0 | 1 | 0 | 46 | 46000 | * | |
4 | 0 | 0 | 1 | 1 | 50 | 50005 | * | * |
Table 7.7 Analytical and simulation results of the TC probabilities () (Example 7.16).
Equivalent stochastic system | Algorithm (Section 7.9.2.3 ) | Simulation | ||||
P | (%) | (%) | (%) | (%) | (%) | (%) |
1 | 1.96 | 1.07 | 1.57 | 0.98 | 1.57 0.26 | 0.82 0.07 |
2 | 2.78 | 1.52 | 2.18 | 1.37 | 2.17 0.17 | 1.17 0.12 |
3 | 3.76 | 2.05 | 2.92 | 1.83 | 2.97 0.21 | 1.53 0.14 |
4 | 4.90 | 2.66 | 3.77 | 2.35 | 4.07 0.27 | 1.93 0.22 |
5 | 6.19 | 3.34 | 4.73 | 2.91 | 4.96 0.38 | 2.36 0.21 |
6 | 7.63 | 4.09 | 5.78 | 3.50 | 5.96 0.31 | 2.95 0.37 |
Table 7.8 Analytical and simulation results of the TC probabilities () (Example 7.16).
Equivalent stochastic system | Algorithm (Section 7.9.2.3 ) | Simulation | ||||
P | (%) | (%) | (%) | (%) | (%) | (%) |
1 | 4.02 | 2.23 | 3.70 | 2.06 | 3.00 0.19 | 1.77 0.08 |
2 | 5.89 | 3.25 | 5.45 | 3.00 | 4.27 0.38 | 2.67 0.12 |
3 | 8.18 | 4.48 | 7.61 | 4.13 | 6.11 0.40 | 3.79 0.28 |
4 | 10.86 | 5.92 | 10.15 | 5.45 | 8.06 0.74 | 5.20 0.27 |
5 | 13.87 | 7.52 | 13.00 | 6.93 | 10.03 0.81 | 6.83 0.44 |
6 | 17.12 | 9.27 | 16.12 | 8.55 | 12.56 0.76 | 8.27 0.26 |
Table 7.9 Analytical and simulation results of the TC probabilities () (Example 7.16).
Equivalent stochastic system | Algorithm (Section 7.9.2.3 ) | Simulation | ||||
P | (%) | (%) | (%) | (%) | (%) | (%) |
1 | 4.62 | 2.57 | 4.41 | 2.43 | 3.74 0.21 | 2.17 0.09 |
2 | 6.82 | 3.76 | 6.56 | 3.58 | 5.18 0.26 | 3.40 0.16 |
3 | 9.50 | 5.21 | 9.20 | 4.97 | 7.54 0.43 | 4.62 0.30 |
4 | 12.62 | 6.88 | 12.29 | 6.60 | 9.72 0.77 | 6.31 0.32 |
5 | 16.09 | 8.76 | 15.74 | 8.43 | 11.78 0.56 | 8.40 0.55 |
6 | 19.80 | 10.78 | 19.45 | 10.43 | 14.03 0.86 | 10.49 0.62 |
In the f‐CDTM under the BR policy (f‐CDTM/BR), b.u. are reserved to benefit calls of all other service‐classes apart from service‐class k. The application of the BR policy in the f‐CDTM/BR is similar to that of the f‐MTM/BR.
Based on the Roberts method and the algorithm of Section 7.8.2.1 of the f‐MTM/BR, the unnormalized values of the link occupancy distribution, , in the f‐CDTM/BR can be determined as follows:
Based on (7.70) and (2.58), we can determine the following performance measures:
Since the finite multirate retry‐threshold loss models are a combination of the retry‐threshold loss models of Chapter 2 and the EnMLM of Chapter , the interested reader may refer to Sections 2.11 and 6.5 for possible applications.
Similar to the previous section, the interested reader may refer to the corresponding section of Chapter 2 (Section 2.12) and Chapter 6 (Section 6.6). In addition to these sections, interesting extensions of the f‐CDTM have been proposed in [7,8], for WCDMA networks. Compared to [ 7], in [ 8], a CAC distinguishes handover traffic from new traffic. More precisely, when the cell load is above a predefined threshold, handover calls are allowed to reduce their bandwidth requirements in order to avoid blocking.