The decision to use Ethernet as a method of transporting audio is well established at this time. However, little discussion has been given to the way audio can impact the IT network. These effects will be our focus in this column. After voice, video was the first media type to be placed on the network. At first, getting it to be acceptable in the IT department took some time as it was not well understood. Large amounts of audio on the network may face the same challenge.
There are three broad categories of Ethernet audio: Dante, AVB, and the others. Dante is a licensed technology that uses the IP packet format, unicasting, and multicasting. Its proponents state that it follows well-established IP standards and doesn’t require special Layer 2 switches. These positions are true. However, from the perspective of the IT department, there are nevertheless some unanswered questions. What are the criteria for using VLANs in Dante networks? Using a VLAN to isolate the audio traffic is generally possible. But what requirements are needed in the routing structure to allow remote control or configuration? What should the guidelines be for the use of QOS relative to other time-sensitive traffic like voice, video surveillance, or manufacturing control? Also, if typical LAN monitoring tools such as those from Netscout, tcpdump, Solarwinds, or Wireshark are used, what information would be evident? Will a network support person get the same detail that they probably could get about an IPTV flow?
[Byte-Sized Lesson in AV/IP: Packet Structures]
AVB is a Layer 2 method of delivery. That is, the packet format must be that which is specified in the IEEE 802.3 standards. However, IP is optional. The composition of the data field is entirely up to the manufacturer who is following the AVB standard. AVB requires switches that implement the standard but not all Layer 2 switches do that. The same questions about VLANs and remote access can be posed about AVB. When the IT tools mentioned above are used, all that will be apparent is that there are many raw Ethernet frames on the network. No information about the audio will be apparent.
One of the confusing points in the discussion of AVB and Dante is about the issue of multicasting. Dante uses Layer 3 multicasting. It uses the 224.x.x.x address scheme and should require IGMP snooping in the switches if the flow is one source to many output devices. On the other hand, according to Daniel Reed at Biamp, AVB uses Layer 2 multicasting. This technology has been available in Ethernet since its inception, but it has only recently become popular. Often it is used in data centers. Here, it is referred to as a fabric network. One advantage to using Layer 2 multicasting is that in the event of a network link disruption, the recreation of correct paths (called reconvergence) is likely to be quicker than it is with Layer 3 multicasting.
