2960 problems with unicast VoD movies

Unanswered Question
Aug 31st, 2009


does anybody expirience problems with VoD unicast packets on Cisco 2960 or 3750? I have situation where I have 2960 lanlite access switches where all IPTV devices are connected. No other devices sharing switch with IPTV devices. 2960s are connected to the 6509 core switch with 1Gbps fiber connections. Video server is connected to 6509 switch. From that video server some VoD movies are played via unicast and some via mcast. It is all in one VLAN, one subnet. On TV mcast traffic is good, without problems. But unicast VoD traffic is making problems in way that very often some packets are droped by NIC card of IPTV box, not by network devices. When you connect IPTV directly to 6509 there are NO problems with unicast traffic. Using wireshark I noticed the difference between traffic pattern of mcast comparing to ucast (attachment) on the port where IPTV is connected. I also expirienced the same problem with 3750 switch. On graphs attached to this message seems like there is no difference between ucast pattern on 6509 and 2960, and still with 2960 there are problems with pixelizations. When you ping IPTV device from 6509 switch at the same time pixelization occures, replay packet for ping request is not recived by 6509. Like the buffer on IPTV device's NIC card droped UDP packet that carries the stream and could not replay to ping request. And all of that because the ucast traffic is more intensive, according to wireshark graphs.

Does anybody has some idea about this problem? Thanks.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
irisrios Fri, 09/04/2009 - 12:59

It can cause problems, especially if it is constantly out of order. You can increase buffers of the IPTV viewers and try.

ivanbarkic Sat, 09/05/2009 - 00:34

I tried that and it didn't help. after noticing that the interruption on video occurs in almost the same intervals and after measuring those intervals we saw that it happens almost every 10 sec. Since the keepalives on Cisco switches for Ethernet is every 10 sec by default, we disabled keepalives on access port of 2960 and the problem was gone. Not one error on video. So, I'm not sure why keepalives are so problematic in this case, specially when it works on 6509 GE copper port with keepalives. Is it the difference in hardware between 6509 GE port and 2960 FE port?


This Discussion