1. Problem review
Jiangxi Unicom's video conference in the province, the provincial main venue to send auxiliary stream, all the sub-sites see the auxiliary stream are suddenly stuck, the mainstream is basically normal.
2. Site networking and version information
[Device and version information]
VP9660 V500R002C10SPCg00
TE40 V600R006C10
SMC V600R006C10SPC800
3. Problem handling and location process
3.1. Video device side positioning analysis
1. From the nlog analysis, the MCU receives the secondary stream normally without packet loss, and the secondary site secondary stream receives normal. It is initially determined that there is a packet loss in the direction of the MCU to the sub-site.
2. Packet capture analysis on the MCU side switch (5300) and the branch site terminal side switch (3328):
The MCU side sends the packet normally without packet loss.
There is a packet loss on the terminal side of the conference site.
There is a packet loss in the preliminary demarcation network.
3. The MCU side has one-touch trace positioning, and the MCU has no internal packet loss.
3.2. Network side positioning analysis
The packet capture analysis shows that there is no packet loss on the 5300 uplink port, and packet loss occurs at the NE40E port.
1) If the NE40E-B loses the seq number to 35522
2) On the S5300 uplink, I caught this package.
After the optical module between the 5300 switch and the NE is replaced, the packet loss problem still exists.
3. Both the 5300 switch and the NE router have developed that they have no problems and cannot be further delimited. After discussion, there is another 5300 switch (B) between the 5300 and the NE router. At the entrance on the 5300-A, the entrance of the 5300-B switch and the NE40 entry are captured.
Simplified networking: MCU (134.237.32.244) à (drop packet inflow) - S5300-Aà (drop packet outflow) - S5300-B (temporary string switch) à NE40E-Bà NE40---A à conference terminal (134.237. 32.11).
Capture analysis:
1) If the S5300-connects to the MCU's entrance to capture packets, it shows no packet loss
2) If the S5300-A is connected to the S5300-A, the UDP packet is found to be 1.76% (continuous packet loss). The result of the NE400E-B packet capture analysis is the same.
3) Packets are also lost at the NE40 entry, and after the RTP packet is compared and analyzed, the packets lost by the NE40 are the same as those of the S5300-A. This confirms that the burst packet loss occurs on the 5300 uplink port.
4. The switch research and development went to the site and found that the upstream optical port of the switch has internal packet loss. (I did not find it before, the 5300 old version v200r001c00spc300 statistics have asked
5. After the 5300 cache is tuned, the problem does not reappear. The switch R&D analysis suspects that the bursty traffic causes the internal cache to be insufficient and the packet loss occurs (no specific data is given).
Description:
Cache of the S5300EI device
The default is 0.36MB.
The highest version of the S5300 is V200R005.
Cache of the S5320EI device:
The default is 1.24MB and the enhanced mode can be increased to 2.2MB. The S5320EI supports the mainstream version V200R011.
Problem root cause analysis summary and solution
[Problem Root]
The 5300 optical port exceeds the interface cache and causes packet loss.
[Solution]
After upgrading the V5R5 version and the latest patch, the S5300 increases the cache and observes that no packet loss occurs.
