Problem Description

1, Optix OSN2500 IPTV system service flow through the transmission equipment is bound to a155M bandwidth Ethernet channel , after the on-demand program, the scene switching faster screen will appear mosaic, the set-top box has the LOS (frame loss) count increase, about 5 to 10 minutes will appear once; a set of program average bandwidth required is 2Mbps.

2. The network diagram is shown below:

Alarm message

The video-on-demand service carried by the SSN4EGS4 board appears to be mosaic.

Processing

1, configure the central office # 210 network element 5 - EGS4-1 to # 213 network element EGS2-1, the central office # 210 network element 4 - EGS4-3 to # 213 network element EGS2-2 two binding 1VC4 bandwidth Ethernet channel, in # 213 network element EGS2 board using fiber optic PORT1 and PORT2 direct loopback, in # 210 network element EGS4 board GE port using industry-recognized SmartBits meter to send 140Mbps rate test (the speed is close to the limit of the binding bandwidth), continuous hanging meter for 58 minutes without any packet loss. The GE port of the EGS4 board of the #210 network element uses the industry-recognized SmartBits meter to send a 140Mbps rate test (the speed is close to the limit of the bound bandwidth ), and there is no packet loss for 58 minutes in a row , proving that there is no problem with the transmission equipment.

2, access to IPTV on-demand business flow, video receiving equipment can connect to the video server, but on-demand program, has been a black screen, analyzed and confirmed to be a video screen receiving equipment operating mode settings, match the operating mode, can be a normal on-demand program, the scene switching faster screen will appear mosaic, set-top box LOS (frame loss) count increases, about 5 to 10 minutes will occur once. From the analysis of the RMON statistics of the GE port of the EGS4 single board, the EGS4 single board does not receive ultra-long packets, ultra-short packets or FCS error packets, and the possibility of error packets is ruled out.

After enabling the flow control of #210 network element 5-EGS4-1 port (docking port with Cisco router), the effect is significantly better, the LOS (frame loss) count of the set-top box is 0 for most of the programs on-demand, and the LOS (frame loss) count of the set-top box is 2 for the scene of playing Fast & Furious 4 with 15-minute screen switching at a fast speed; and after prohibiting the flow control, the LOS (frame loss) count of the set-top box is increased by 5 times. , indicating that there are serious bursts of IPTV service streams. The judgment is based on the fact that the service chip of the EGS4 board has an anti-pressure mechanism, if the egress cache is full, the service chip will automatically send a Pause frame to the upstream device (i.e., the Cisco device) through the inbound port, and the upstream device should pause sending the service message after receiving the Pause frame, so that the EGS4 board can avoid discarding the packet because of its inability to cache bursty packets.

4. Configure GE port mirroring in CISCO6509, use ETHREAL packet capture tool to analyze it, disable flow control on port 5-EGS4-1, and start both packet capture and VOD operations.

VOD "Fast and Furious 4 " scene changes very fast and change a lot of the film, observe the set-top box LOS (frame loss) count changes and mosaic time point, found that the 228th second, the set-top box LOS (frame loss) count increased by 3, to analyze the captured data.

5. Analyze the captured packets of the Fast & Furious 4 program and find that there is a serious burst of packets when the mosaic occurs in the 228th second, and the captured data is as follows:

The 83145th frame timestamp is 228.716956S

Frame 83186 timestamped at 228.718349S

The duration of these 42 frames is only 0.001393 seconds, these 42 frames total about 37.2Kbytes, the bandwidth burst in this time window reaches: 37.2*1000*8bytes/0.001393 seconds = 213.6Mbps.

Analyzing the video data stream, the average flow of the program is only 2Mbps, in each Ethernet frame length of an average of 888bytes (packet capture statistics obtained), 1 second only need to send 281.5 frames, in the emergence of the mosaic, only 0.001393 seconds (1/717 seconds ) continued to send out 42 frames , which42 frames, if the average rate of sending, it takes 42/281.5 seconds, i.e., about 1/7 seconds to complete. This 42 frames, if sent at an average rate, would take 42/281.5 seconds, i.e., about 1/7th of a second to complete. Therefore, the results from the packet capture analysis show that there is a serious traffic burst in this video service stream.

6. As the bandwidth convergence from GE to 1VC4 channel is realized by using the transmission Ethernet board, there is a bandwidth bottleneck, as shown in the following figure:

Area B is a cache that can be used by the single board for traffic bursts. When the size of the A-area portion of the burst traffic exceeds that of the B-area, normal packet loss occurs. As long as bandwidth convergence exists, normal packet loss may exist in the case of severe traffic bursts. (As a transmission device, the port cache queue of the EGS4 single board is small, with only 40Kbytes per port, in addition to the additional overhead field inside the single board, the actual cached messages can only reach 20Kbyte at least, so the tolerated burst messages are also small, and the instantaneous burst messages are prone to cause normal packet loss when there is a bandwidth convergence.)

7. In order to further verify the conflict of traffic burst on bandwidth convergence, the following network is used

Without transmission equipment, use Cisco Gigabit switches to complete the GE to FE convergence and then further test to verify whether the switch can withstand the burst of service flows:

It is verified that the video signal will appear mosaic phenomenon when the program on demand test is conducted under this networking, which indicates that the cache of this switch is also insufficient to withstand the impact of the burst traffic of this video service. Therefore, the serious bursting of the video source traffic is the key to the problem.

8, customer feedback W city to H city channel is only one155M bandwidth, but no mosaic problem, after analyzing and confirming that the business flow is using the TPC/IP protocol, the business flow through the transmission of binding 1VC4 bandwidth Ethernet channel bearer, on-demand business does not have any mosaic phenomenon, because the TPC/IP protocol has a retransmission mechanism, which also provides a proven solution for the subsequent solution. a validated solution.

9. Since the customer's video equipment does not support the TPC/IP protocol for transmitting video streams and the customer is not willing to turn on the upstream router traffic shaping function, the increase of 1x speed bandwidth is used, which greatly alleviates the probability of packet loss, and the customer's on-demand video streams are no longer mosaicked.

1. Packet loss occurs in the transmission Ethernet channel;

2. There are severe bursts in the Optix OSN2500 IPTV on-demand stream packet text.

Recommendations

1. Upstream routers complete traffic shaping based on VLAN service flows;

2、Video source equipment to complete a better instantaneous traffic regulation control , from the source to solve the problem;

3、Because the Ethernet board configuration GE port to bind aVC4VCTRUNK there will be a bandwidth bottleneck, increase the bandwidth to alleviate the probability of packet loss;

4, when there is a serious burst of traffic, resulting in a small number of packet loss, IPTV uses the UDP protocol does not have a retransmission mechanism, such as the use of TCP/IP protocol, you can circumvent the problem, but also to circumvent the mosaic due to optical path errors caused by the problem.