If you're just running VoIP and some run-of-the mill data traffic, coming up with a marking scheme for your QoS...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
isn't all that complicated. You can generally get away with assigning all your voice packets an IP Precedence of 5, which means "Critical" and send everything else with the default value of 0, otherwise known as "Routine". But, what values should you choose if you also have Video over IP or some data applications that are more important than others?
Unless you've got a good reason to deviate, go with the following general guidelines:
At layer 2, RTP (Real Time Transport Protocol) traffic should be assigned a value of 5. At layer 3, give it an IP Precedence of 5 or a DSCP value of "Expedited Forwarding" (EF).
Assign your VOIP control traffic (which includes things like skinny, MGCP, H.225.0, and H.245) a value of 3 for the L2 CoS, and 3 for IP Precedence. For DSCP, use AF31. This means "Assured Forwarding Class 3 with a low drop probability". You can read more about the Assured Forwarding classes in RFC2597 which defines them.
For videoconferencing traffic, assign it a value of 4 at layer 2 and IP Precedence. For DSCP, use a value of AF41, which is class 4, and still a low drop probability.
For streaming video, use a value of 1 in both the IP Precedence and layer 2 CoS. For DSCP, use AF13, which is class one, with a high drop probability, as this is less preferred and also more tolerant of dropped packets than most of the previous types of traffic.
Finally, for your data, it is generally recommended that you use values 0 to 2 as appropriate, for layer 2 CoS and layer 3's IP Precedence. For DSCP, use AF13 or AF23, which are classes one and two, with a high drop probability. If your data traffic is non-reliable (e.g. something besides TCP-based) and not very tolerant of packet loss, you can change the drop probability to medium by using AF22.
Keep in mind that while this is a good strategy to implement, it is likely that you still have devices in your network that don't yet support the level of granularity offered by DSCP. In other words, it may treat classes 1 and 2 the same, or may ignore your drop probability. As always, ask your vendor what they support, and then test it to be sure.
Thomas Alexander Lancaster IV is a consultant and author with over ten years experience in the networking industry, focused on Internet infrastructure.