CC BY
11
FORMATION OF EQUIVALENT SIMULATION MODEL OF AN REALTIME VIDEO STREAM GENERATOR USED IN PACKET-ORIENTED COMMUNICATION NETWORKS, TAKING INTO ACCOUNT THE STRUCTURE OF THE H.264 COMPRESSION ALGORITHM
DOI 10.24411/2072-8735-2018-10325
Mikhail M. Lukichev,
St.Petersburg state transport university
of the emperor Alexander I, St.Petersburg, Russia,
Keywords: Telecommunication Networks, the simulation model, probabilistic-temporal characteristics of video traffic, delay time variation, traffic-sniffer, video traffic H.264.
At present time, the need for transmitting video traffic in packet-oriented communication networks with a set quality is growing rapidly. If the telecommunication network does not meet the requirements of video devices, then digital distortions and even loss of video frames occur. However, assessing the requirements for communication networks is a complex, multi-criteria task, especially at the stage of network design. The most rational way to assess the behavior of a communication network is to build a simulation model. This model allows you to define the performance indicators of the telecommunications network for each load flow passing through it. Such indicators include: delays, delay deviations, packet loss, and distortion [1]. Delay deviation is one of the most difficult to formalize parameters, but its role in high-quality video image transmission is very important. Accurate definition of the delay deviation, when building simulation models, it is not enough to use the known laws of the distribution of time between IP packets, because this approach does not take into account the structure of the real-time video stream. To solve the problem of accurately building an equivalent real-time video traffic generator, we analyzed the parameters of time intervals between the moments of receiving IP packets received from a network video camera. The analysis showed that the probability density function of the time intervals between neighboring IP packets, in this example, consists of seven independent functions, as well as five additional functions that form the structure of the video stream. Based on the analysis, we constructed an algorithm for the interaction of the considered generators, each of which is based on the unique parameters of various distribution functions, which together form the equivalent real-time video traffic. Also, in the simulation environment, a simulation model is built that has the characteristics of the studied traffic obtained experimentally. The resulting model more accurately displays the structure of video traffic and can be used as a tool for analyzing the properties of a video stream during its aggregation and movement over a telecommunication network. In addition, the approach to the formation of an equivalent network load generator considered in the article can be presented as a methodical for constructing traffic generators from various sources, allowing you the best accuracy at describing the properties of the object under study.
Information about author:
Mikhail M. Lukichev, St.Petersburg state transport university of the emperor Alexander I, Postgraduate of the department "Electrical Communication", St.Petersburg, Russia
Для цитирования:
Лукичев ММ. Формирование имитационной модели эквивалентного генератора видеопотока реального времени, используемого в пакетно-ориентированных сетях связи, с учетом структуры алгоритма сжатия H.264 // T-Comm: Телекоммуникации и транспорт. 2019. Том 13. №11. С. 43-52.
For citation:
Lukichev M.M. (2019). Formation of equivalent simulation model of an real_time video stream generator used in packet-oriented communication networks, taking into account the structure of the H.264 compression algorithm. T-Comm, vol. 13, no.11, pр. 43-52.
Introduction
Video transmission In IP-oriented telecommunication networks hold a special place. A significant increase in video content has forced public networks to increase bandwidth to transfer large amounts of user data. A large number of protocols have been developed for both encoding video data and transmitting it to a large number of users (multicast). These trends have led to a quality transition in security and video surveillance networks. At the moment, it is considered the most promising deployment and/or implementation of video surveillance networks based on packet data transmission. Analog video surveillance based on channel switching is economically justified only for a small number of cameras. Typically, the number of analog video cameras on an object does not exceed 16, and ihe length of each line is not more than 500 m relative to the central video processing station.
Video systems that allow continuous monitoring of objects with the ability to store video data are very popular. They provide passive protection of various objects, and also allow you to quickly determine the sequence and chronology of events in the investigation of non-standard situations. Therefore, video surveillance systems built on the basis of packet switching (IP) technology are especially relevant. At the moment, the number of installed cameras is increasing, which leads to the need to increase the performance of data networks.
However, the widespread installation of video cameras leads to increased requirements for data transmission networks. Practice shows that determining the exact values of these requirements is a complex multicriteria task, the optimal solution of which depends on the particular object under consideration.
Unlike other types of traffic, video traffic has the following features:
• A large amount of information transmitted per unit time
* Strict requirements for network quality parameters (errors, delay variation, packet loss, etc.)
In order to reduce the load on the data network, the H.264 codec was developed. It can significantly reduce network bandwidth requirements by compressing video data. This approach allows us to simplify the storage of video data, i.e. reduce the amount of space required for writing to the hard disk. However, the reproduction of such a video stream requires a lot of processing power from the video display station.
The relevance of the study is to determine the characteristics of the video stream, which are the source data for the formation of requirements for telecommunication networks.
Object of research
1. Video resolution 1920xl080r (FullIID).
2. Frame rate - 20 frames / sec.
3. [-frame interval - 20.
4. MTU- 1500.
5. ConstantBitRate.
6. Maximum BitRate-8192 Kbps.
7. H.264 video codec.
Also, a video surveillance workstation was connected to the same switch, on which display of the video stream and traffic sniffer were simultaneously launched [2|. The image that entered the camera's lens was static and unchanged during the study period. The 25 minutes research was sufficient to identify the characteristics of video traffic. During the indicated time, 1,047,586 RTP packets were received, which were encapsulated in UDP packets, then in IPv4 and Ethernet as indicated in Fig. 2.
OJ
c t o_ CL
0J JZ > Û- □ 1— CÉ H.264
+J LU
Fig. 2. The structure of the IP-packet video stream
During the analysis of time intervals between successively arriving IP packets to the workstation interface using the classical statistics approach [3] (Sturges rule), we can obtain the histogram shown in Fig. 3, while the average lime interval between packets is 0.0013 seconds, and the average frame duration is 1458 bytes, which corresponds to a bit rate of approximately ((1458 +14) * 8/0.0013) = 8.63 Mbit/s.
-5-moKi
■.oxooo
8X000 630000 30JUUU 230000
s s 2 a
a s s s
H ? ^ 3
i i s a ;sj
=> => o o o
rv r*
K s
s s s
s
r-ÖÖÖÖÖÖÖÖÖäÖÖÖO
$ a s £
S » S
5 S § Î
" a s ä
3 a s s
5 3 8 S
fl Q N "î
N * kl N
3 5 S
IK «A
m u*>
s »
s s
Fig. 3. The histogram of the distribution of the time interval between the packets of the studied video stream eonstructcd according to the Sturges rule
Let's take a look at the structure of video traffic. To do so, the model is shown in Fig. 1
Fig, I. Installation diagram for characteristics research of the H.264 video traffic
A network camera with the following settings was connected to the switch with POE support:
However, this approach docs not reflect ail the characteristics of the ongoing processes, the structure of traffic and network mechanisms. When you change the scale of the timeline, as well as filtering time intervals, you can get some mathematic patterns. To accurately determine the temporal characteristics of the video stream, it is necessary to determine the traffic structure. To do this, we will determine how video encoding occurs using the H.264 codec [4]. The structure of the video stream (in relation to the initial data indicated above) can take the form shown in Fig. 4, Taking into account the initial data, we get that every 1 second it is necessary to transfer the full picture that fell into the camcorder's lens.
W
20 (rame pwiod (every 20th frame - It
l-fr*me P-frame
LU
Lengthj
Defeat Wtj
P-franw B-frame
y u...
Fig. 4. The structure of the researched video stream, using the H.246 codec
This is followed by a P-frame containing information about the image change in 1/20 sec, relative to the I-frame. Then the B-frame contains information about the image change for the same 1/20 sec, but relative to both the P-frame and I frame. This approach to encoding a video stream allows you to reduce the requirements for the bandwidth of the communication channel, and also allows you to increase the "depth" of the archive of video data. However, when playing such a stream, increased demands are placed on the performance of the video playback device, because it is necessary to store the data of 1 and P frames and con-
stantly calculate the changes in the frame relative to the indicated frames.
Thus, in order to obtain a complete model of the delay generator that takes into account the structure of video traffic, from the obtained time series consisting of the moments time of registering the reception of IP packets on the interface of the studied video surveillance station, it is necessary to identify the boundaries for all frames I, P and B, to evaluate the time delays between the indicated frames, to evaluate the duration of each, and also determine the time intervals between adjacent images within each frame. To determine the start and end of a frame, you can use the Timestamp, which indicates the absolute time that belongs to each frame. Since the information contained in one image cannot fit in one IP packet, several such packets transmit information about one frame. The genera! view of the data obtained by the sniffer is shown in Fig. 5. It can be seen from the figure that the allocation of different frames can be used service "Mark", according to the recommendation of RFC 3984 [5J.
After highlighting the frame boundaries of the video stream, you need to determine what type of frames these packets belong to. At the first stage, the boundaries of the beginning, the end of each frame were determined. The total number of all frames selected in this way was 29268. Then, based on the source data, there should be 1464 I-frames and 27822 P / B-frames in the analyzed traffic. Accordingly, traffic capture lasted 1464/60 = 24.4 minutes, which corresponds to the source data.
Tune Source Oestnabon PrctDrol Lenglh
772 21.165049 192.*VX.**.54 192.**X:**,57 RTP 1482
773 21.165242 192. ™x,*x, 54 192. xxx. xx. 57 OTP 1482
774 21.165243 192. ***.**. 54 192.***;**. 57 RTP 1374
775 21.204290 192.**X.**.54 192.***:**. 57 RTP 1482
776 21.204515 192 .■■xxx.xx. 54 192,***i**, 57 RTP 1482
777 21.284721 192. **x.*x. 54 192.***:**. 57 RTP 1482
778 21.204722 192, »**.•**, 54 192, **X;-X*. 57 RTP 1432
779 21.204928 192. xxx.xx. S4 192,***^**, 57 RTP 1432
730 21,205133 192.™«:»*. 54 192.***;**, 57 RTP 1482
781 21.205133 192.**X:-XX.54 192. *»*;**. 57 RTP 1482
732 21.205339 192. *KX.x x. 54 192 .xxx.xx. S7 «TP 1432
783 21.205549 192.™*.**. 54 192.***:**, 57 RTP 1482
734 21.205550 192.**x.*x. 54 192.***:**. 57 RTP 1482
785 21.205745 192. xx. 54 192 xxx.xx. S7 RTP 1482
736 21.205950 192. *xii:XX.54 192.***.**. 57 RTP 1482
787 21.205950 192, ■***;■**. 54 192.**X;**. 57 RTP 1482
738 21.206153 192,w*.*x+S4 192. xxx. xx. S, 7 RTP 1432
789 21.206362 192.***,**.S4 192 .***:**, 57 RTP 1482
790 21.206363 192.**X;*x. 54 192. *»*:**. S7 RTP 830
791 21.245046 192,wx,wx+54 192.*WX;*)t, 57 RTP 1432
792 21.245247 192.***.**, 54 192. *»*:**, 57 RTP 1482
793 21.245437 192.r**x.x*. 54 192.***;**. 57 RTP 1482
794 21.245627 192.wx.wx.S4 192 .xxx.xx. 57 RTP 1482
795 21.245627 192.***.**.S4 192.**X:*K, 57 RTP 1482
796 21.245819 192.**x.*x. S4 192. *»*:**. 57 RTP 1482
797 21.246010 192, yKx.x* S4 192. xxx.XX. 57 RTP 1432
798 21.246011 192. **X;*K.54 192. *»*:**, 57 RTP 1482
799 21.246201 192.***.**. 54 192. **X; **. 57 RTP 1482
800 21.246391 192. yxx. xx. S4 192. xxx.x'x. 57 RTP 1432
301 21.246572 192. **x.*x,54 192.**Xi**, 57 RTP 1482
302 21.246573 192.***.**. 54 192. *»*:**. 57 RTP 1482
303 21.246774 192, WX7XX, 54 192.WiXiW0.57 RTP 1432
804 21.246775 192,***.**. S4 272.xxx xx. 57 RTP 998
805 21.285729 192.***.**. 54 192.***;**. 57 RTP 1482
806 21.235938 . «XX. XX. Ç.4 192. XKX.XX. 57 RTP 1432
307 21.286131 192. ***:**. 54 192.*w*i**v 57 RTP 1482
808 21.286132 192 .***-.**, 54 192.**X;*K. 57 RTP 1482
809 21.236323 192,W0X.XX,54 192.'X*X;XX, 57 RTP 1432
310 21.236539 192, ***:**■.54 192.**XT*K. 57 RTP 1482
Sll 21.236539 192. *»*:**. S4 192. *x*,**, 57 RTP 1482
312 21.236735 192,*»X;-ïPX<54 192.xxx.xx. 57 RTP 1432
813 21.286929 192. ***.**. 54 192. **x.**, 57 RTP 1482
B14 21.287122 192.***:**. 54 192.***.**. 57 RTP 1482
815 21,237123 192, wx.-**,54 1&2 .xxx.xx. 57 RTP 1432
Wb
PT-OynamicRTP PT-DynemicRTP PT-DynanicRTP PT.=DynamicRTP PT.0yn4micRTP PT-OynamicRTP PT-OynamicRTP PT-Dynamic RTP PT-DynamicRTP PT»DynafflicRTP PT-DynamicRTP PT=DynamicRTP PT-DynamicRTP PT- Dynamic RTP PT-DynamicRTP PT-OynaeiicRTP PT-DynamicRTP PT-DynemicRTP PT-OynamicRTP Pr.DynamicRTP PT-OynamicRTP PT-OynamicRTP PT. Dynamic RTP PT-OynamicRTP PT-OynamicRTP PT-OynamicRTP PT-DynamicRTP PT=0ynaraicRTP PT«DynamicRTP PT-DynamicRTP PT-OynamicRTP PT.DynamicRTP PT-DynamicRTP PT-DynamicRTP PT-DynamicRTP PT-DynamicRTP PT-DynamicRTP PT»DynamicRTP PT-DynamicRTP PT-DynamicRTP PT-DynamicRTP PT-OynamicRTP PT-DynamicRTP PT-DynamicRTP
■Type-96, -Type-96, -Type-96, "Type-96, -Type-96, -Type-96, -Type-96, ■Type-96, -Type-96, ■Type-96j ■Type-96, -Type-96, -Type-96, ■Type-96, -Type-96, -Type-96j ■Type-96, -Type-96, ■Type-96, ■Type-96, -Type-96, -Type-96j ■Type-96, -Type-96, -Type-96, ■Type-96, -Type-96, -Type-96j ■Type-96, ■Type-96, -Type-96, ■Type-96, ■Type-96, ■Type-96, ■Type-96, -Type-96, ■Type-96, ■Type-96, -Type-96, -Type-96 j ■Type-96, -Type-96, -Type-96, ■Type-96,
SSRC-0X48D29 SSRC-0X48Û29 SSRC-0X4BD29 SSRC=0x4BD29 5SRC=0x4BD29 SSRC-0X4BD29 5SRO0X48O29 SSRC-BX48D29 SSRC-0X48D29 SSRC-0X4B029 SSRO0X48D29 SSRC=0x4BO29 SSRC=0x4SD29 55RC-0X48029 SSRC-0X4BD29 SSRC-0x48029 SSRC-0X48O29 SSRC=Bx4BD29 SSRC-0X48O29 SSRC-0X48O29 SSRC=0x4BO29 SSRC=0x4BO29 5SRC-0X48O29 SSRC-0X4BD29 SSRC-0X4BO29 55RC-0X48D29 SSRC-0X4SD29 SSRC=0x4BD29 55RC-0X48029 SSRC-0X48O29 S5RC-0X4BO29 5SRC-0X48D29 SSRC-0X48D29 SSRC-0X4SO29 5SRC-0X48D29 SSRC=0x48O29 SSRC-0X4SO29 55RC-0x48029 SSRC-BX48D29 SSRO0X4SD29 SSRC-0X4SO29 SSRC-0X4BO29 SSRC-0X4BD29 SSRC-0X4SO29
Seq-11641 Seq-H842 Seq-11B43 Seq =11344 Ssq=11345 Seq=llB46 Seq-U347 Seq-11848 Seq-llS49 Seq-IISSB Seq»1185i Seq=11852 Seq-11853 Seq-11854 Seq-11S55 Seq=11856 5eq-11357 Seq=11B58 Saq-11359 Seq-11860 Seq=11861 Seq=11662 Seq-11363
Seq-11364 Seq-11869 Seq >11866 Seq=11367 Seq=11366 Seq-11869 Seq-11870 Seq-11871 Seq-11872 Seq-11873 Seq-11874 Seq-11375 Seq=11376 Seq-11S77 Seq-11078 Seq-11379 Seq-11380 Seq-116S1 Seq-11332 Seq-11BB3 Seq-11834
Time* TimeTime. T i rv T irn."-Time* T ime-Time. TimeTime. TimeTime* Timc time« Time-T irv-Time« Time* Tin»= Time. Time= T irvc:
Time. TimeTime. TimeTime. Time= TimeTime. Time. TimeTime. TimeTime-Time= Time. TimeTimeTime. TimeTime. Tiiw-Time-
37334359
37834850
37834850, Ha rit
37838450
37838450
3783S45B
37838450
37838458
378384SB
37835450
37838450
37838450
3733S4SB
37858458
3733B45B
37338450
37838450
37S3B450
37338450, Hark
37842650
37842050
37842050
37842050
37842050
373420S0
37842050
37842050
37842050
37642050
37842050
37842050
37942050
37842050, Hark
37845650
37845650
37845650
37845650
3784S6S0
37845650
37845650
37845650
37845650
37845650
37845650
Fig. 5. The part of data received by the traffic sniffer (Dump-data)
Number of packets in ¡-frames
Number of packets in P/B-frames
300 250
zoo
150 100 50 0
ll
I I
7000 6C00 5000 4000 3COO 2COO 1COO 0
I I
219 220 221 222 22J 214 225 226 227 22S 229 14 16 IS 30 22 24 26 28 30 32 34 36 38 40 42 44
Fig. 6. Distribution histogram of the of the number of IP packets in I and P/B frames
Next, the number of IP-packets included in the frames was checked. Analysis of the number of IP packets in one frame helped to identity from the whole set of I-frames. Exactly 1464 frames had a significantly larger number of IP packets (more than 200), while in the remaining frames the number of frames did not exceed 100. Figure 6 shows histograms of the number of packets included in the I-frame and P/B-frames based on which the frame is differentiated.
At the same time, a clear boundary separating the frames P from B, according to the criterion of the number of packets contained in each of them, could not be found. This is explained by the "scene" of video recording, as the picture that fell into the lens review was relatively stable (static), then the P and B frames were identical in the number of packets included in their composition. With increased amount of movement entering the camera's lens, differences in the durations of frames P and B should appear. To obtain the simplest distribution laws, at this stage we will focus on a "static" scene.
Distribution of time intervals between IP packets
"inside" a frame
At the next stage, regularities were revealed in the time intervals between packets in I-frames, as well as in P/B-frames. During the analysis, it turned out that the probability-time characteristics of the intervals between consecutive arrival of IP packets in frames I, P and B are almost identical, but a significant differ-
ence was observed in the duration parameters of the frames under consideration, as well as inter-frame time intervals. Therefore, it is advisable to make an identical law of the distribution of time between packets within any frame. However, it is not possible to apply the Sturges rule in view of the multimodality of the distribution of time intervals. In the course of data analysis, the distribution densities of the time intervals between IP packets in I-frames and separately in P/B-frames were obtained. Two main time intervals were identified, in which the absolute majority of the studied time intervals lit (more than 96%). The remaining time intervals were for the duration after i-frames (about 0.14%) and alter P/B - frames (about 2,6%), which will be considered later. The histogram of the distribution of intervals between IP packets is shown in Fig, 7.
As can be seen from Fig. 7, the distribution density of time intervals between IP packets is independent of the type of frame. The histogram shapes of the time intervals between IP packets belonging to I and P/B frames are the same. In order to, to simplify the construction of the model of the video traffic generator, these distributions can be considered as independent of the type of frame. It should be noted that the emergence of multimodality of the distributions of the considered time intervals is due to limitations in the bit rate in the settings of the camera. Thus, at a certain initial moment of time, the IP packet intensity increases sharply (see range 1 in Fig, 7).
Range 2 [0-8 [isec.] The share IP-packets in range 1 of the total is 35%
JflOOW
îaoao îhhm ma» s«« îwœo saxo
WXP «MO
MECO
lll.ll.....
¡•Ï ps*f
■ diïinixjiwn in P/Bfrwn« (23H)
■ dHinbutkwi in I ífJf!» (UK)
Range 2 [176-230 psec.] The share IP-packets in range 2 of the total is 60%
lili
II
IklLlLl
11111 h § i § § g 11 s i 11 s 111 i 11111 i s ! 111g s 11111111 s 11 i i § 11111 § i § i
o o" o o o o o o o oooaooaod o o o o" d a o o o o o" o o o' o o" o o o a o o o o o o o o a o o a o a o
* both distribution I63H) • distribuí ion In P/ft frames 14M) ■ distribution in I tram» Í21H)
Fig. 7* Histograms of the time distribution of intervals between IP-packets inside frames I and P/B
By the form of the histogram of the distribution density of time intervals after I-lrames, we assume that its probability density function obeys the Gamma law; and we also determine his parameters: a = 1.91, and (i = 0.007. Then the root mean square error will be 31%, which is sufficient results due to the significantly (by several orders of magnitude) smaller volume of intervals created by the considered function, relative to the volume of other traffic.
Distribution of time intervals after P/B-frames
After we inspected time intervals after P and 13 frames, their difference became obvious. The shape of the histogram of the distribution density of time intervals after P and B - frames was the same, and the mathematical expectations are different. Accordingly, for each time interval after P and after the B frame, histograms of the distribution density and probability functions presented in Fig. 13 and 14, respectively.
In both cases, to the probability density function of the distribution of time intervals after B-frames, the Beta distribution with the parameters presented in Table 1 was used. 2. Moreover, the root mean square error does not exceed 8% for each distribution.
0,1 R o,>G 0,14 0,12 0,1 0,08 0,0b 0,0-1 0,02
0,16 0,14 0.12 0,1 0,08 o,oc
0,04 0,02
\
y \
i k
_/ X
A 1
1 1 1 l>M
<S> o" C>> Qj' QY cjV qV <jY sV <jY QjV (JV
Fig. 14. A histogram of the distribution density of time intervals after B-frames and a graph of the probability density distribution
Table 2
Parameters probability density functions of the distribution of time intervals after the P/B frame
/
A
Function for the interval after a ß A (mill) B(max)
P-frame 15,402 386 0,03521 0,1
B-frame 8 1049 0,05631 0,26
CO CC OC CO CC CO 00
ÂSîftrSîiiîïirïl&îïîmnRiSrSpSriSiSSîSîSl
Fig, 13. A histogram of the distribution density of time intervals after P-frames and a graph of the probability density distribution
Thus, after evaluating all the time interframe intervals that make up the video traffic structure, we can conclude about the complexity, heterogeneity and multimodal ity of the resulting stream. After analyzing the time intervals of the structural elements of the video traffic, for each element we can identify the main distribution laws and their parameters, presented in table 3.
For a brief indication of the distribution of time intervals under consideration, we used the abbreviations shown in the diagram in Fig. 4.
Table 3
Parameters of the probability density functions of the distribution of time intervals for cach structural element of the video stream
Designation Physical meaning Percentage of total intervals Distribution law Distribution Function Parameters
DJnJ D_in_P D_in_B Time intervals between consecutive packets arriving at the time of transmission of ¡/P/B frames 97% Normal 1 M = 1,49*10"; STD = 1,15*10""
Normal 2,1 M = 1,81 * 10"1; STD = 2*10"''
Normal 2.2 M = 1,9* 10"4; STD = 2*10"
Normal 2.3 M = 1,98*10 ; STD = 0,0000021
Normal 2,4 M= 2,063* I0"1; STD = 2* 10*"
Normal 2.5 M = 2,131*10 ; STD = 1,8* lO"6
Norma! 2.6 M = 2,183* 10'4; STD = 5*10'"
DAfterl Time intervals between receiving the last packet belonging to the I-frame and the first to the P-frame 0,14% Gamma a =1,91 ß = 0.007
DAfterP Intervals between receiving the last packet belonging to a P/B frame and the first frame following it 2,66% Beta a = 15,402; [3 = 386 A (min)= 0,03521; B {max}= 0,1
DAfterB Beta a = 8; p = 1049 A {niin)= 0,05631; B{max)=0,26
Length! Intervals between receipt of the first and last packet belonging to the l-frame n/a Beta et ■= 4,9601 ; ß = 418,94 A (min> 0,0261; B (max)= 1,2204
LengthP Length B Time intervals between receiving the first and last packet belonging to a P/B frame n/a Normal M = 3.79* 10"'; STD = 5,2 *10~l
References
This approach allows us to achieve accurate results in the calculations and modeling of complex video systems. The resulting model of an equivalent video stream generator based on the 11.264 codec can be repeatedly used to create an equivalent aggregated stream created by several sources and can be used for further researches.
When integrating the obtained model of generators into a simulation model of the functioning of telecommunication communication networks, it is possible to evaluate the quality parameters of the functioning of such a network described in [ 11, Unlike the known models, the presented model will allow us to more accurately determine the transmission delay parameter, as well as variations in the transmission delay (jitter) that occur during the transmission of video traffic in telecommunication networks.
1. ITU-T Recommendation Y.1541 (12/2011). Network performance objectives for 11'-based services.
2. Getman A.I., Evstropov E.F., Markin Yu.V. (2015). Real-time network traffic analysis: a review of applied tasks, approaches and solutions, Preprint JSP RAS 28, pp. I -52.
3. Vadzinsky R.N. (2001 ). Handbook of Probabilistic Distributions. St. Petersburg: Nauka, pp. 295.
4. ITU-T Recommendation 11.264 (06/19) Advanced video coding for generic audiovisual services.
5. RFC 3984 (February 2005) RTI* Payload Format for H.264 Video.
6. Kleinrock L. (1970). Communication networks (stochastic message flows and delays). Moscow: Nauka, pp. 256.
7. Aliev T.I., Nikulsky I.E., Pyatayev V.O. (2009). Modeling and analysis of the aggregation level of a multiservice télécommunications network. Communication Engineering, vol. 2, pp. 12-18.
8. Boev V. D. (201 1). A study of the adequacy ofGPSS World and Any Logic in modeling discrete-event processes'. Monograph. St. Petersburg: VAS, pp. 404.
ФОРМИРОВАНИЕ ИМИТАЦИОННОЙ МОДЕЛИ ЭКВИВАЛЕНТНОГО ГЕНЕРАТОРА ВИДЕОПОТОКА РЕАЛЬНОГО ВРЕМЕНИ, ИСПОЛЬЗУЕМОГО В ПАКЕТНО-ОРИЕНТИРОВАННЫХ СЕТЯХ СВЯЗИ,
С УЧЕТОМ СТРУКТУРЫ АЛГОРИТМА СЖАТИЯ H.264
Лукичев Михаил Михайлович, Петербургский государственный университет путей сообщения Императора Александра I,
Санкт-Петербург, Россия, [email protected]
Аннотация
в настоящее время существенно возрастает потребность в передачи видео трафика в пакетно-ориентированных сетях связи с заданным качеством. Если телекоммуникационная сеть не удовлетворяет требованиям видеоустройств, то возникают цифровые искажения, и даже потри изображений. Однако оценить требования к сетям связи является сложной, многокритериальной задачей, особенно на этапе проектировании сети. Наиболее рациональный способ оценки поведения сети связи - это построение имитационной модели. Имитационная модель позволяет определить показатели качества функционирования телекоммуникационной сети для каждого проходящего по ней потока нагрузки. К таким показателям относят: задержки, девиации задержек, потери пакетов и искажения. Девиация задержки один из самых сложно формализуемых параметров, однако его роль в качественной передаче видео изображения очень высока. Для точного определения девиации задержек, при построении имитационных моделей недостаточно использовать известные законы распределения времени между IP-пакетами, т.к. такой подход не учитывает структуру видео потока реального времени. Для решения задачи точного построения эквивалентного генератора видео трафика реального времени был проведен анализ параметров временных интервалов между моментами получения IP-пакетов, получаемых от сетевой видеокамеры. Произведенный анализ показал, что функция распределения плотности вероятности временных интервалов между соседними IP-пакетами, в данном примере состоит из семи независимых функций, а также из пяти дополнительных, формирующих структуру видео потока. На основании произведенного анализа построен алгоритм взаимодействия рассмотренных генераторов, каждый из которых основан на уникальных параметрах различных функций распределения, в совокупности формирующий эквивалентный видео трафик реального времени. Также в среде имитационного моделирования построена имитационная модель, которая обладает характеристиками исследуемого трафика, полученного экспериментальным путем. Полученная модель более точно отображает структуру видео трафика, может быть использована как инструмент для анализа свойств видеопотока при его агрегации и перемещении по телекоммуникационной сети. Кроме этого, рассмотренный в статье подход к формированию эквивалентного генератора сетевой нагрузки может быть представлен как методика формирования генераторов трафика от различного рода источников, позволяя наиболее точно описывать свойства исследуемого объекта.
Ключевые слова: сети связи, имитационное моделирование, вероятностно-временные характеристики видео трафика, вариация времени задержки, трафик-сниффер, видео-трафик H.264.
Литература
1. ITU-T Recommendation Y.1541 (12/2011) Network performance objectives for IP-based services.
2. Гетьман А.И., Евстропов Е.Ф., Маркин Ю.В. Анализ сетевого трафика в режиме реального времени: обзор прикладных задач, подходов и решений. Препринт ИСП РАН 28, 2015. С. 1-52.
3. Вадзинский Р.Н. Справочник по вероятностным распределениям. Спб.: Наука, 2001. 295 с.
4. ITU-T Recommendation H.264 (06/19) Advanced video coding for generic audiovisual services.
5. RFC 3984 (February 2005) RTP Payload Format for H.264 Video.
6. Клейнрок Л. Коммуникационные сети (стохастические потоки и задержки сообщений). М.: Наука, 1970. 256 с.
7. Алиев Т.И., Никульский И.Е., Пяттаев В.О. Моделирование и анализ подуровня агрегирования мультисервисной телекоммуникационной сети. Техника связи, 2009. № 2. С. 12-18.
8. Боев В.Д. Исследование адекватности GPSS World и AnyLogic при моделировании дискретно-событийных процессов: Монография. СПб.: ВАС, 2011. 404 с.
Информация об авторе:
Лукичев Михаил Михайлович, аспирант, Петербургский государственный университет путей сообщения Императора Александра Санкт-Петербург, Россия
