Você está na página 1de 24

Models of Systems with Multi-rate

Overflow Traffic
Chapter 8

Maciej Stasiak, Mariusz Głąbowski


Arkadiusz Wiśniewski, Piotr Zwierzykowski
Overflow one-rate systems

A R, D (s2,Z)
...

• Parameters of overflow traffic:


• Mean value of overflow traffic R,
• Variance of overflow traffic s2,
• Peakedness (variance/mean) Z = s2 / R
• Spreadness (variance-mean) D = s2 - R

2
Overflow one-rate systems

A Offered traffic

t
Parameter A Y R
y Carried traffic

Peakedness
Z=1 Z<1 Z>1
factor t

R Overflow traffic
Spreadness
D=0 D<0 D>0
factor
t

3
Overflow one-rate systems
• Mean value and variance of overflow traffic:

R  AEV ( A) R  AEV ( A)

 A   A 
s  R
2
1 R D  R  R
V 1 A  R  V 1 A  R 

• For offered traffic (Poisson type) mean value is equal to


variance:

R  A, s 2  A, D  0

4
ERT (Equivalent Random Traffic) method
- step 1

A1 Ak Overflow traffic parameters:

V1 Vk Ri  Ai EVi ( Ai )

  Ai 
 i
D  R 
i  R 
i
R1, D1 Rk, Dk  Vi  1  Ai  Ri 

k
R, D R   Ri
i 1

Vov D   Di
i 1

5
ERT (Equivalent Random Traffic) method
- step 2

equivalent group

A1 Ak A*

V1 Vk V*

R1, D1 R1, Dk R, D

Vov Vov

Traffic stream overflowing from equivalent group is the same as


the sum of traffic streams overflowing from real groups

6
ERT (Equivalent Random Traffic) method
- step 2

equivalent group

A1 Ak A*
V1 Vk V*

R1, D1 R1, Dk R, D

Vov Vov

 s 2 s 2 
 A*  s  3   1
2

 R  R  Rapp approximate solution



 (R2  s 2 )
V *  A * R 2  s 2  R  R  1

7
ERT (Equivalent Random Traffic) method
- step 3
• Blocking probability of total offered traffic in the system

A*

V*
E  E(V *V o b) ( A*)
R
A * E(V *V ob) ( A*)
Vov Bob 
R
A*E(V* Vov )(A* )

8
Hayward’s equivalence method

 Ri  Ai EVi ( Ai )
A1 Ak 
  Ai 
V1 Vk  Di  Ri  V  1  A  R  Ri 
  i i i 

R1 Rk k k
R   Ri , s   s i2 ,
2
Z  R /s 2
i 1 i 1
Vov

Z s 2 / R R
E  EVo v ( )
Z
Z

9
Overflow multi-rate systems – MODEL I
• Assumptions:
o Only one class of traffic streams is offered to each of primary
groups of the system
o There are M primary groups in the system
o the group i has capacity equal to Vi BBU
o Each primary group is offered different traffic stream
The primary group i is offered traffic with intensity Ai
o Class i call requires ti BBU to set up connection

10
Overflow multi-rate systems – MODEL I

A1,t1 A2,t2 AM,tM

V1 V2 VM

R1, s21 R2, s22 RM, s2M

V
E1, E2, …, EM

11
Overflow multi-rate systems – MODEL I
• Blocking probabilities in the primary groups are
determined by:
• Erlang formula
o one call of class i occupies simultaneously ti BBUs – from the
perpective of Erlang formula it is equivalent to be on the
decrease ti -times in real group
A
capacity Vi BBUs:
i

E  EVi / ti  Ai 

• Kaufman-Roberts formula
o only one stream of calls with demands ti BBUs is offered to the
group with capacity Vi BBUs:
 
Vi
n Pn  Vi  Ai ti Pn ti Vi
E  P  n Vi
n Vi ti 1

12
Overflow multi-rate systems – MODEL I
• Blocking probabilities of primary groups are the basis for
determining of overflow traffic parameters:
o Mean value Ri
o variance si2
o Peakedness factor Zi

13
Overflow multi-rate systems – MODEL I

Occupancy distribution of alternative group:

Blocking probability: Z=?

14
Overflow multi-rate systems – MODEL I
• Equivalent peakedness factor determination:

• Equivalent peakedness factor is directly proportional to


peakedness factors of individual traffic classes.

15
Dimensioning algorithm for MODEL I
(alternative group with M classes)
• Algorithm

1. Determination of the so-called component group capacities


(V1, V2, …, VM ) under the assumption that only one traffic
stream is offered to the alternative group:

2. Determination of the initial value V/Z for alternative group


on the basis of values Vi/Zi calculated in the previous step:

16
Dimensioning algorithm for MODEL I
(alternative group with M classes)
• Algorithm

3. Serially decreasing in capacity of component groups as long as the


following conditions are fulfilled:

where:

KR designates Kaufman-Roberts model.

17
Dimensioning algorithm for MODEL I
(alternative group with M classes)
• Algorithm

4. Determination of alternative group capacity by summing up all


component capacities V1+V2+...+VM (obtained in the previous
step).

18
Overflow multi-rate systems – MODEL II
• Assumptions:
o There are M primary groups in the system
o Each primary group is offered M classes of traffic streams
o traffic intensity of class i offered to the primary group j is equal
to the value Ai,j Erl.,
o Class i call requires ti BBU to set up connection

19
Overflow multi-rate systems – MODEL II
A11, A12, …, A1M A21, A22, …, A2M AK1, AK2, …, AKM
t1, t2, …, tM t1, t2, …, tM t1, t2, …, tM

V1 V2 VM

R11, R12, …, R1M R21, R22, …, R2M RK1, RK2, …, RKM


s211, s212, …, s21M s221, s222, …, s22M s2K1, s2K2, …, s2KM

V
E1, E2, …, EM

20
Overflow multi-rate systems – MODEL II
• Algorithm
1. Blocking probability Ei,j of class i traffic stream in the primary group j is
determined by Kaufman- Roberts formulae:
M
n P(n)   Ai , j ti P(n  ti )
i 1
Vj

Ei , j   P ( n)
n V ti 1

2. Blocking probabilities calculated in the previous step are the basis for
determining mean values of overflow traffic streams Ri,j:

21
Overflow multi-rate systems – MODEL II
3. Decomposition of each primary group into M fictitious component groups
with capacities Vij. Each fictitious component group services only given
class calls. It makes possible to use Riordan formulae for determination
of the variance σi,j of class i traffic stream overflowing from primary
group j.

A1, j Ai, j AM, j A1, j Ai, j AM, j

?
Vj V1, j Vi, j VM, j

R1, j Ri, j RM, j R1, j Ri, j RM, j


s1, j si, j sM, j s1, j si, j sM, j

22
Overflow multi-rate systems – MODEL II

4. Determination of fictitious component Free


group capacity Vij : resources fictitious
class i group
occupancy

where: Yi,j – average number of class i


calls serviced in the group j (class i occupancy i
occupancy):

23
Overflow multi-rate systems – MODEL II

5. Calculation of the variance sij2 of individual traffic streams overflowing


from fictitious groups :

where quotient Vi,j/ti normalize the system to the one-rate case.

6. Mean value and variance of traffic streams offered to the alternative


group:

7. To evaluate the value of the capacity in the alternative group the


dimensioning algorithm described in the MODEL I can be used.
24

Você também pode gostar