Professional Documents
Culture Documents
I. INTRODUCTION
CA
BCP
VASP
GW
Management Transport
Quality
control
Reliable
Unreliable
transport prot.
transport prot.
Network prot.
streaming data
other data
charging information
CDR transfer
ME
RNS
NodeB
RNC
CN
SGSN
C
G
F
GGSN
FW Proxy
BS
OCS
I PBBPBBP I
EN
Other
SP
VASP
1-p
1-q
1
q
The Gilbert model has only two states with the following
probabilities: 0 represents the event that a packet is
received successfully while 1 refers to the opposite. p
denotes the probability the packet is lost, provided that the
next is not lost, q is the opposite. 1-q is called conditional
loss probability (clp). If p+q=1 the Gilbert model reduced
to the Bernoulli model.
The state probabilities can be calculated in the following
way:
q
p
(1)
S0
, S1
pq
pq
p and q can be computed from the streaming data flow,
using the loss length distribution statistics [7]:
f
i 1
m0
m (i 1) ,
i
, 1 q
(2)
i 1
mi
i
i 1
p(n-2)(n-1)
1
p10
n-1
p(n-1)0
p(n-1)(n-1)
0
p23
0
#
#
#
0
0
0
1
0
0
0
#
0
We can compute the state probabilities like this
S0 S0
S S n 1
1
1 , S 1.
T
i
# #
i 0
S ( n 1) S ( n 1)
The probability distribution of loss runs (pk)
n 1
p01
n 1
m0 , p( k 1)( k )
i 1
(4)
(5)
n 1
m m .
i
i 1
(6)
i k 1
1
408
544,40
402,32
411,45
V. CONCLUSION
We have presented a new, streaming proxy based
charging architecture for QoS-differentiated charging in 3G
networks. The recommended solution is available for online
and offline charging. We have pointed out the importance of
bursty packet loss, which determines the quality of
streaming, in contrast with the average loss rate. We have
proposed several methods for QoS-differentiated charging,
which are in compliance with the recommendations of
3GPP. We have described three models, which could predict
the QoS of a streaming flow. The algorithms were evaluated
them on a real GPRS streaming flow.
In the future we would implement our charging method in
a simulation environment and we plan make an
experimental charging system, which can be used to
evaluate the developed methods.
REFERENCES
[1]
[2]
[3]
[4]
[5]
4
5
0,00
1,29
5,04
1
0,00
0,19
1,01
[6]
[7]
[8]
[9]