I have a 10mbs link between 2 sites. One end has a Cat6513 Hybrid and the other a Cat3560. I am running Voice traffic (SIP) across this link
I am trunking between these two sites. I am using auto qos in both switches, using trust cos on trunked links and trust dscp on non-trunked links.
When I do pacture captures with Etheral I'm seeing all my UDP/RTP voice traffic as 'ef' and some other non UDP/RTP traffic as 'cs2'
I want to prioritize SQL traffic across this link so it is higher then default dscp of 0, but still below voice traffic priorities. I need to do this because one of my applications that talks to the SQL server keeps freezing.
To do this, I setup a policy map on the 3560 that matches tcp/1433 and defined this class is as dscp 28, which is af32. I applied this inbound on interface where computers are connected that use the application in question.
When I do a packet capture on the 6513,s panning the port that connects to this remote site, I'm seeing SQL traffic from the computers showing a dscp of af31. This is where I get confused. If the switch is set to tag this as af32, why is it coming across as af31? The ISP is not re-tagging any traffic. They are doing Q-in-Q tunneling (encapsulating my encapsulated packet, sending it across, then removing the extra encapsulation, leaving my original packet).
Do I have something setup improperly? The reason I'm concerned is because I typically see QoS examples for voice traffic matching af31 and ef for RTP. Since I'm not seeing af31 on my RTP traffic, but ef, I assume it's not a huge issue, but I really do not fully understand the priority queues and how and their precedence. The goal is to have highest priority to all my voice traffic, then give priority to SQL traffic, then give everything else default 0.