JP5126015B2 - Packet distribution quality evaluation apparatus and method - Google Patents

Description

  The present invention relates to a quality evaluation apparatus and method for packet distribution.

  The transmission quality of real-time services such as voice and video is, for example, that the voice is ITU-T recommendation G.264. 107 and the R value defined in the TTC standard JJ-201.01, the video is ITU-T recommendation JJ. It is evaluated by the packet transmission quality measurement defined in H.241.

  In addition, a method for evaluating the quality of service by placing a quality measurement agent on a SIP server or a video distribution server and measuring server load (for example, CPU usage rate or memory usage rate) or application response has been established.

In relation to the above, it is known to measure the traffic fluctuation of the RTP packet by counting the number of packets for each target server at a constant period and at a constant period.
Japanese Patent Laid-Open No. 2001-144771 Japanese Patent No. 3611823

  However, the above evaluation method has a problem that the quality of RTP packet delivery cannot be evaluated in real time and with high accuracy.

A distribution quality evaluation apparatus according to an embodiment includes a characteristic information acquisition unit that determines a temporal pattern of a traffic amount of distribution data received from a distribution server via a provider network, and the temporal pattern obtained by the characteristic information acquisition unit. based on the change, have a abnormality detecting unit for detecting an abnormal condition in the distribution server or the provider network, the change of the temporal pattern of the traffic volume is the change of the period of the distribution data, the abnormality detecting unit Specifies whether the abnormal condition is in the distribution server or in the provider network .

The distribution quality evaluation method according to an embodiment includes a step of receiving distribution data transmitted from a distribution server via a provider network, a step of obtaining a temporal pattern of the traffic volume of the distribution data, and the temporal pattern. Detecting an abnormal state in the distribution server or the provider network based on the change in the distribution amount, and the change in the temporal pattern of the traffic volume is a change in the period of the distribution data, and the abnormal state is detected. The step identifies whether the abnormal condition is in the distribution server or in the provider network.

  The quality of RTP packet delivery can be evaluated in real time and with high accuracy.

  Embodiments will be described in detail with reference to the drawings. Throughout the drawings, the same or corresponding components are denoted by the same or corresponding reference numerals.

  Hereinafter, the case of video distribution will be described as an example, but the present invention can also be applied to distribution of other real-time data such as audio distribution.

  FIG. 1 is a diagram illustrating a video distribution network 100 according to an embodiment. The video distribution network 100 includes a provider network (MVNO / SNI) 110 including video distribution servers 1 to 3, a packet copy device 111, and a quality management system 112. The video distribution network 100 includes a core network 120 including core network routers 1 to 4. Further, the video distribution network 100 includes an access network 130 including video receiving terminals 1 to 3 such as personal computers and IP televisions.

  For example, the RTP packet is transmitted from the video distribution server 1 of the provider network 110, routed by the core network routers 1 to 4 of the core network 120, and received by the video receiving terminals 1 to 3 of the access network 130.

  The RTP packet is copied by a packet copy device 111 (for example, TAP) of the provider network 110, and the copied RTP packet is sent to the quality management system 112.

  FIG. 2 is a block diagram illustrating a quality management system 112 according to one embodiment. The RTP packet copied by the packet copy device 111 is sent to the packet collection unit 1121 that collects packets flowing through the network in the quality management system 112.

  The RTP packet collected by the packet collection unit 1121 is sent to the header information analysis unit 1122 that detects the head of the frame from the header information and cuts out a frame for one period.

  The frame for one period cut out by the header information analysis unit 1122 is sent to the characteristic information management unit 1123 that calculates the characteristic value in the cut out frame. The characteristic value is, for example, the number of peaks and valleys of traffic included in one frame, the interval between peaks and valleys, and the height difference between peaks and valleys.

  The characteristic value calculated by the characteristic information management unit 1123 is sent to the quality deterioration detection unit 1124 that notifies the user of an alarm when the characteristic value exceeds a threshold value.

  The quality management system 112 further includes a database 1125. In one embodiment, the characteristic information management unit 1123 stores the calculated characteristic value in the database 1125, and the quality degradation detection unit 1124 accesses the database 1125, and the characteristic value exceeds the threshold based on the stored characteristic value. The user may be notified of an alarm when In another embodiment, the characteristic information management unit 1123 stores the calculated characteristic value in the database 1125, accesses the database 1125 as necessary, reads out the stored characteristic value, and sends it to the quality degradation detection unit 1124. Also good.

  As described above, the distribution quality evaluation apparatus (quality management system 112) according to the above embodiment transmits a provider network (for example, the provider network 110 in FIG. 1) from a distribution server (for example, the video distribution servers 1 to 3 in FIG. 1). A characteristic information acquisition unit (for example, characteristic information management unit 1123) for obtaining a temporal pattern of the traffic amount of the distribution data (for example, RTP packet) received via the network. Further, the distribution quality evaluation apparatus is configured to detect an abnormal state in the distribution server or provider network (for example, an encoding process in the distribution server, an IO card abnormality, a provider network An abnormality detection unit (for example, a quality degradation detection unit 1124) that detects a network abnormality such as congestion.

  FIG. 3 is a diagram illustrating a data structure of the distribution server information table 300 indicating the characteristic values stored in the database 1125. FIG. 4 is a diagram for explaining the characteristic values.

  The distribution server information table 300 includes an IP address table 301 indicating the IP addresses of the video distribution servers 1 to 3, a mountain valley number table 302 associated with the IP address table 301, a mountain valley interval table 303, and a mountain valley high and low width table 304. including.

  As shown in FIG. 4A, the peak / valley number table 302 indicates the number of peaks 41 and valleys 42 in the traffic amount of one frame for each distribution server.

  As shown in FIG. 4B, the mountain / valley interval table 303 indicates intervals (ad) for each pair of peaks and valleys (indicated by the numbers between the valleys) in the traffic amount of one frame for each distribution server.

  As shown in FIG. 4C, the mountain valley height table 304 shows the height for each distribution server for each mountain and valley (indicated by the valley number) in one frame traffic volume.

  As described above, the distribution quality of the RTP packet can be detected by measuring the characteristic value of the frame for one period and examining the change in the pattern for each distribution server.

  FIG. 5 is a diagram for explaining the separation of abnormal points in the provider network. FIG. 5A shows a normal case, FIG. 5B shows a case where the distribution server has an abnormality, and FIG. 5C shows a case where the provider network other than the distribution server has an abnormality.

  When the provider network is normal, it is known that the traffic volume changes at a constant period as shown in FIG.

  On the other hand, when an abnormality occurs in the distribution server, it is known that a certain delay occurs in the time required for one cycle, as shown in FIG. This will be described in detail later with reference to FIG.

  Further, it is known that when an abnormality occurs in the network, a random delay occurs in the time required for one cycle as shown in FIG. This is because the magnitude of the delay changes due to another service load (interrupt of other service packets).

  FIG. 6 is a diagram for explaining in more detail a case where an abnormality has occurred in the distribution server (for example, distribution server 1 in FIG. 1). 6A shows a normal case, FIG. 6B shows a case where an abnormality occurs in the IO card of the distribution server, and FIG. 6C shows a case where an abnormality occurs in the encoding processing unit of the distribution server. Indicates.

  When the distribution server is normal, as shown in FIG. 6A, the traffic amount of the video distribution data indicated by reference numeral 61 is sent out by an IO card having the performance indicated by reference numeral 62. Therefore, the quality management system 112 The data received by (see FIG. 1) is as indicated by reference numeral 63.

  When an abnormality occurs in the IO card of the distribution server, as shown in FIG. 6B, the traffic volume of the video distribution data indicated by reference numeral 61 is indicated by reference numeral 64 (throughput is lower than performance 62). Since the data is transmitted by an IO card having performance, the data received by the quality management system 112 (see FIG. 1) is as indicated by reference numeral 65. The received data indicated by reference numeral 65 has a longer period, a greater number of peaks and valleys, and a lower height of the peaks and valleys than normal received data indicated by reference numeral 63.

  Conversely, when the period of the received data received by the quality management system 112 becomes longer, the number of peaks and valleys increases, and the height of peaks and valleys decreases, it is considered that an abnormality has occurred in the IO card of the distribution server. be able to.

  When an abnormality occurs in the encoding processing unit of the distribution server, as shown in FIG. 6C, the traffic amount of the video distribution data indicated by reference numeral 66 (a processing delay occurs due to congestion or the like and the peak decreases) However, the data received by the quality control system 112 (see FIG. 1) is as indicated by reference numeral 67. The reception data indicated by reference numeral 67 has a longer period than the normal reception data indicated by reference numeral 63.

  In other words, when the period of the reception data received by the quality management system 112 is long, it can be considered that an abnormality has occurred in the encoding processing unit of the distribution server.

  FIG. 7 is a flowchart illustrating a quality evaluation method 70 for RTP packet delivery according to one embodiment. The quality evaluation method 70 can be executed by, for example, the quality management system 112 shown in FIG.

  In step S71, the quality evaluation method 70 collects packets flowing through the network. For packet collection, for example, a packet copy device (for example, the packet copy device 111 shown in FIG. 2) is inserted into the network, and the packet copied by the packet copy device is received.

  In step S73 following step S71, the beginning of the frame is detected from the header information of the collected packets, and a frame for one cycle is cut out. This can be determined, for example, by referring to a packet unit start indicator (PUSI) included in header information of a transport stream (TS) packet. Then, the time until the next top packet is captured can be determined as one cycle.

  In step S75 following step S73, a characteristic value is calculated. The characteristic value is, for example, the number of peaks and valleys of traffic volume in one cycle, the interval between peaks and valleys, and the height of each valley.

  In step S77 following step S75, it is determined whether an interval of one cycle is extended. If not (NO), the process proceeds to step S79 and it is determined that the packet delivery is normal.

  On the other hand, if the interval of one cycle is extended in step S77 (YES), the process proceeds to step S81.

  In step S81, it is determined whether the interval of one cycle is not constant but extends randomly. If it is randomly extended (YES), the process proceeds to step S83, and it is determined that there is an abnormality in the provider network (network). This is as described with reference to FIG.

  On the other hand, if the method of extending the one-cycle interval is not random in step S81 (NO), the process proceeds to step S85.

  In step S85, it is determined whether only the time until the first burst point is large. The first data transmission in one cycle has a characteristic of bursting because the encoded data is transmitted using the maximum performance of the IO card. When only the time to the first burst point (section a in FIG. 4B) becomes large, there is a possibility that an abnormality has occurred in the encoding processing unit in the video distribution server. On the other hand, when there are changes in the intervals and configuration widths of the sections b, c, and d in FIG. 4B, the performance of the IO card may be degraded. Furthermore, when the maximum height of each mountain and valley is small, the performance of the IO card may be deteriorated.

  Therefore, in step S85, when only the time to the first burst point is large (YES), the process proceeds to step S87, and it is determined that there is an abnormality in the encoding processing unit of the distribution server.

  On the other hand, if the time other than the time until the first burst point is larger in step S85 (NO), the process proceeds to step S89 and it is determined that there is an abnormality in the IO card of the distribution server.

  In another embodiment, an abnormality in a component of a provider network (provider network 110 in FIG. 1) including a video distribution server (for example, video distribution servers 1 to 3 in FIG. 1) and reception of a received packet of the quality management system 112 Based on the causal relationship other than those described above between the patterns, an abnormality in the component of the provider network (provider network 110 in FIG. 1) may be determined based on the reception pattern of the received packet of the quality management system 112.

  As described above, the distribution quality evaluation method according to an embodiment receives the distribution data sent from the distribution server via the provider network (step S71), and the temporal pattern of the distribution data traffic volume (for example, (Steps S73 and S75), and detecting an abnormal state in the distribution server or the provider network based on the temporal pattern change (steps S77 to S89). ).

  FIG. 8 is a diagram for explaining an abnormality determination method using characteristic values. The quality deterioration detection unit 1124 illustrated in FIG. 2 can create the characteristic management table 800 illustrated in FIG. 8 based on the characteristic values received from the characteristic information management unit 1123.

  The characteristic management table 800 has areas 801 to 803 for each distribution server.

  In the area 801, the number of peaks and valleys, a threshold value specified in advance by calculation or an operator, and possible causes when the number of peaks and valleys exceeds the threshold value are entered.

  Further, in the area 802, the valley number, the interval between the valleys, the threshold value specified in advance by calculation or the operator, and possible causes when the interval between the valleys exceeds the threshold value are entered. Yes.

  In the area 803, the valley number, the height of the valley, the threshold value specified in advance by the operator or the operator, and the possible cause when the height of the valley exceeds the threshold value are entered. .

  The cause shown in each area | region 801-803 can set an arbitrary message for the user for every characteristic of a delivery server.

  For example, in the distribution server 2, the interval between the valley-to-valley numbers 2 is 40 and exceeds the corresponding threshold 30, so it is considered that an encoding abnormality has occurred. As described with reference to FIG. 2, the quality degradation detection unit 1124 can notify the user of this cause as an alarm.

  In the provider network (for example, the provider network 110 in FIG. 1), a load balancer may be arranged in order to balance the load on each distribution server. FIG. 9 is a diagram showing a provider network 910 in which a load balancer is arranged.

  The provider network 910 is the same as the provider network 110 shown in FIG. 1 in that it includes the video distribution servers 1 to 3 and the packet copy device 111, but distributes additions to the video distribution servers 1 to 3. The difference is that a load balancer 911 and a quality management system 912 are included.

  In the provider network 110, the video distribution servers 1 to 3 have the same IP address and are distinguished by the port number of the load balancer 911. Therefore, the quality management system 912 manages the video distribution servers 1 to 3 based on not only the IP address but also the port number of the load balancer 911. Accordingly, when the video distribution server is load-balanced by the load balancer, the quality management system 912 specifies the video distribution server by specifying the IP address and the port number. The other configuration of the quality management system 912 is the same as that of the quality management system 112 shown in FIG.

  In the above embodiment, the abnormality in the provider network is detected based on the temporal pattern of the received data in the quality management system. In another embodiment, an anomaly in the provider network may be detected based on the probability density of the packet arrival interval of the received data. The temporal pattern of the received data described with reference to FIGS. 5 and 6 is also reflected in the probability density of the packet arrival interval of the received data. Therefore, an abnormality in a distribution server or a network can be detected by obtaining a probability density of a packet arrival interval of received data without necessarily obtaining a temporal pattern of received data.

  FIG. 10 is a diagram illustrating a provider network 1010 according to an embodiment. The provider network 1010 includes a packet copy device 1011, a quality management system 1012, and a network 1013.

  The packet copy device 1011 copies the packets A to D flowing through the network 1013 and sends them to the quality management system 1012.

  The quality management system 1012 obtains the probability density based on the packet arrival interval based on the packet received from the packet copy device 1011.

For example, the packet A, B, C, D is _t each time arriving at the quality control system 1012 _a, t _b, t c, when the _{t d,} the packet arrival interval t _b -t _a
t _c -t _b
t _d -t _c
It becomes. The probability density can be obtained by (number of packets with respect to packet arrival interval) / (total number of packets received by the quality management system during a certain time).

  FIG. 11 is a diagram illustrating a graph 1100 as an example of the probability density according to the packet arrival interval. In the graph 1100, traffic is concentrated when the packet arrival interval is around 0, 7, and 17 seconds. If the data encoding type is the same, the same characteristics are obtained regardless of the content to be distributed.

It will be apparent to those skilled in the art that the distribution quality evaluation apparatus or method according to the above embodiment can solve the following problems associated with the conventional method.
In the case of voice quality measurement or the like, RTP packets are transmitted at a constant cycle of 10 to 20 msec. However, in the case of video distribution, RTP packets are not always transmitted at a constant cycle due to the influence of server load, encoding load, etc., and it is difficult to monitor distribution quality with high accuracy. Although it is possible to evaluate the average value or the total amount of traffic in a long cycle, there is a risk that detection of deterioration in distribution quality may be delayed.
-When analyzing characteristics in a short cycle, it is difficult to evaluate because the average value varies.
・ When analyzing characteristics over a long period, the trend can be leveled, but the analysis takes time, and the results of sudden changes (rapid increase and decrease) are rounded to the average value.
-If the server is deadlocked due to a problem, the quality measurement agent existing on the same system is also affected, and normal evaluation cannot be performed.
There is a problem that the quality measurement agent itself imposes a load on the service server and degrades the service performance.

  Although the embodiment of the present invention has been described in detail above, the present invention is not limited to the specific embodiment, and various modifications and changes are within the scope of the gist of the present invention described in the claims. It can be changed.

The following supplementary notes are described for the above embodiment.
(Appendix 1)
A characteristic information acquisition unit for obtaining a temporal pattern of a traffic amount of distribution data received from a distribution server via a provider network;
A distribution quality evaluation apparatus comprising: an abnormality detection unit that detects an abnormal state in the distribution server or the provider network based on a change in the temporal pattern obtained by the characteristic information acquisition unit.
(Appendix 2)
The change in the temporal pattern of the traffic volume is the period of the distribution data,
The distribution quality evaluation apparatus according to appendix 1, wherein the abnormality detection unit specifies whether the abnormal state is in the distribution server or the provider network.
(Appendix 3)
The change in the temporal pattern of the traffic volume is the time to the first burst point in one period of the distribution data,
The distribution quality evaluation apparatus according to attachment 1 or 2, wherein the abnormality detection unit specifies that the abnormal state is in an encoding process of the distribution server when only the time until the first burst point changes.
(Appendix 4)
The distribution quality evaluation apparatus according to attachment 3, wherein the abnormality detection unit specifies that the abnormal state is in an IO card of the distribution server when a time other than the time until the first burst point changes.
(Appendix 5)
The distribution quality evaluation apparatus according to attachment 1, wherein the change in the temporal pattern of the traffic amount is a probability density pattern with respect to a packet arrival interval.
(Appendix 6)
The distribution quality evaluation apparatus according to any one of appendices 1 to 5, wherein when the distribution server is load-balanced by a load balancer, the distribution server is specified by a port number to which the distribution server is connected.
(Appendix 7)
Receiving distribution data sent from the distribution server via the provider network;
Obtaining a temporal pattern of the traffic volume of the distribution data;
A delivery quality evaluation method comprising: detecting an abnormal state in the delivery server or the provider network based on the change in the temporal pattern.
(Appendix 8)
The change in the temporal pattern of the traffic volume is the period of the distribution data,
The distribution quality evaluation method according to appendix 7, wherein the step of detecting the abnormal state includes a step of specifying whether the abnormal state is in the distribution server or in the provider network.
(Appendix 9)
The change in the temporal pattern of the traffic volume is the time to the first burst point in one period of the distribution data,
The step of specifying the abnormal state includes the step of specifying that the abnormal state is in the encoding process of the distribution server when only the time to the first burst point is changed. Delivery quality evaluation method.
(Appendix 10)
The distribution according to claim 9, wherein the step of identifying the abnormal state includes a step of identifying that the abnormal state is in the IO card of the distribution server when the time other than the time until the first burst point changes. Quality evaluation method.
(Appendix 11)
The distribution quality evaluation method according to appendix 7, wherein the change in the temporal pattern of the traffic volume is a pattern of probability density with respect to a packet arrival interval.
(Appendix 12)
The distribution quality evaluation method according to any one of appendices 7 to 11, wherein when the distribution server is load-balanced by a load balancer, the distribution server is specified by a port number to which the distribution server is connected.

1 is a diagram illustrating a video distribution network according to an embodiment. FIG. 1 is a block diagram illustrating a quality management system 112 according to one embodiment. FIG. It is a figure which shows the data structure of the delivery server information table which shows the characteristic value stored in a database. It is a figure for demonstrating a characteristic value. It is a figure for demonstrating isolation | separation of the abnormal location in a provider network. It is a figure for demonstrating in more detail the case where abnormality arises in a delivery server. It is a flowchart which shows the quality evaluation method of RTP packet delivery by one Embodiment. It is a figure for demonstrating the abnormality determination method by a characteristic value. It is a figure which shows the provider network by which the load balancer is arrange | positioned. 1 is a diagram illustrating a provider network according to one embodiment. FIG. It is a figure which shows an example of the probability density by a packet arrival interval.

Explanation of symbols

100 Video distribution network 110 Provider network 111 Packet copy device 112 Quality management system 120 Core network 130 Access network 300 Distribution server information table 301 IP address table 302 Number of mountains and valleys table 303 Mountain valley interval table 304 Mountain valley high and low width table 911 Load balancer 912 Quality management system 1010 Provider network 1011 Packet copy device 1012 Quality management system 1121 Packet collection unit 1122 Header information analysis unit 1123 Characteristic information management unit 1124 Quality degradation detection unit

Claims (5)

A characteristic information acquisition unit for obtaining a temporal pattern of a traffic amount of distribution data received from a distribution server via a provider network;
Wherein based on a change in the temporal patterns characteristic information obtaining unit is determined, it possesses an abnormality detection unit for detecting an abnormal condition in the distribution server or the provider network,
The change in the temporal pattern of the traffic volume is a change in the period of the distribution data,
The distribution quality evaluation apparatus , wherein the abnormality detection unit specifies whether the abnormal state is in the distribution server or the provider network . The change in the traffic volume of the temporal pattern is the time change until the first burst point in one period of the distribution data,
The distribution quality evaluation apparatus according to claim 1 , wherein the abnormality detection unit specifies that the abnormal state is in an encoding process of the distribution server when only the time until the first burst point changes. The distribution quality evaluation apparatus according to claim 2 , wherein the abnormality detection unit specifies that the abnormal state is in an IO card of the distribution server when a time other than the time until the first burst point changes. The distribution quality evaluation apparatus according to claim 1, wherein the change in the temporal pattern of the traffic amount is a change in a probability density pattern with respect to a packet arrival interval. Receiving distribution data sent from the distribution server via the provider network;
Obtaining a temporal pattern of the traffic volume of the distribution data;
Based on a change in the temporal pattern, seen including a step of detecting an abnormal condition in the distribution server or the provider network,
The change in the temporal pattern of the traffic volume is a change in the period of the distribution data,
The step of detecting the abnormal state is a distribution quality evaluation method that specifies whether the abnormal state is in the distribution server or the provider network .

Images