Introduction to Audio Video Bridging Networks
Information about Audio Video Bridging (AVB)
Audio and video equipment deployments have traditionally been analog single-purpose point-to-point one-way links. Migration to digital transmission also continued to retain the point-to-point one-way links architecture. The dedicated connection model resulted in a mass of cabling in professional and consumer applications, which was hard to manage and operate.
In order to accelerate the adoption to Ethernet based audio/video deployments in an interoperable way IEEE came up with the IEEE Audio Video Bridging standards - IEEE 802.1BA. This defines a mechanism where endpoints and the network will function as a whole to enable high quality A/V streaming across consumer applications to professional audio-video over an Ethernet infrastructure.
Note |
|
Licenses Supporting AVB
AVB is supported on the following two license levels only:
-
ipbase
-
ipservices
Benefits of AVB
AVB is a standard based mechanism to enable Ethernet based audio-video transmission which has the following benefits:
-
Guaranteed max Latency
-
Synchronized Time
-
Guaranteed Bandwidth
-
Professional Grade
Components of AVB Network
AVB protocols operate only in domains where every device is AVB capable. The AVB network comprises of the AVB talkers, AVB listeners, AVB switches and the grandmaster clock source.
-
AVB Talker - An AVB end station that is the source or producer of a stream, i.e. microphones, video camera, and so on.
-
AVB Listener - An AVB end station that is the destination or consumer of a stream, i.e. speaker, video screen, and so on.
-
AVB Switch - An Ethernet switch that complies with IEEE802.1 AVB standards.
-
AVB stream: A data stream associated with a stream reservation compliant with the Stream Reservation Protocol (SRP).
Note
In some instances, the word “bridge” is used. In this context, it references to a switch.
The IEEE 802.1BA specification requires that an AVB talker must be grandmaster capable. In a typical deployment a network node can also be the grandmaster, provided it can either source or derive timing from a grandmaster capable device and provide the timing to the AVB network using IEEE 802.1AS.
Figure 1 shows a simple illustration of AVB network with different components.
In many instances, the Audio/Video end points (Microphone, Speaker, etc.) are analog devices. AVB end-point vendors introduce Digital Signal Processors (DSP) and I/O devices that provide extensive audio/video processing and aggregate the end-points into an AVB Ethernet interface, as shown in Figure 2.
Supported SKUs for AVB
AVB is supported on the following Catalyst 3850 and Catalyst 3650 SKUs.
-
WS-C3650-24PDM
-
WS-C3650-48FQM
-
WS-C3650-8X24UQ
-
WS-C3650-12X48UQ
-
WS-C3850-12X48U
-
WS-C3850-12XS
-
WS-C3850-16XS
-
WS-C3850-24XS
-
WS-C3850-24XU
-
WS-C3850-32XS
-
WS-C3850-48XS
Note |
In Cisco IOS XE Denali 16.3.1, AVB is supported only on the non-mGig interfaces on WS-3850-12X48U. Starting from Cisco IOS XE Denali 16.3.2, AVB is supported on the mGig interfaces on WS-3850-12X48U and WS-C3850-24XU. |
Information about Generalized Precision Time Protocol (gPTP)
Generalized Precision Time Protocol (gPTP) is an IEEE 802.1AS standard, which provides a mechanism to synchronize clocks of the bridges and end point devices in an AVB network. It defines the mechanism to elect the grandmaster clock (BMCA) among the time-aware bridges and talker and listener. The grandmaster is the root of the timing hierarchy that gets established in the time-aware network and distributes time to nodes below to enable synchronization.
Time synchronization also requires determining the link delay and switch delays in the network nodes. The gptp switch is a IEEE 1588 boundary clock, which also determines the link delay using the peer-to-peer delay mechanism. The delays computed are included in the correction field of the PTP messages and relayed to the end-points. The talker and listener use this gPTP time as a shared clock reference, which is used to relay and recover the media clock. gPTP currently defines only domain 0, which is what the switch supports.
The peer to peer delay mechanism runs on STP blocked ports as well. No other PTP messages are sent over blocked ports.
In a PTP domain, Best Master Clock (BMC) algorithm organizes Clocks and Ports into a hierarchical fashion, which includes clocks and port states:
-
Grandmaster (GM/GMC)
-
Boundary Clock (BC)
-
Master (M)
-
Slave (S)
-
Passive (P)
Information about Multiple Stream Reservation Protocol (MSRP)
Multiple Stream Reservation Protocol (MSRP) provides a mechanism for end stations to reserve network resources that will guarantee the transmission and reception of data streams across a network with the requested QoS. It is one of the core protocols required on an AVB device (talker, listener and switches). It allows talkers to advertise streams across a network of AVB switches and listeners to register for receiving the streams.
MSRP is the key software protocol module for supporting AVB. It enables stream establishment and teardown. It interfaces with gPTP to update the latency information for the streams. It interfaces with the QoS module to setup the hardware resources that would guarantee requested bandwidth for the streams. It also provides the QoS shaping parameters required for the credit based shaper.
Note |
When AVB is enabled globally the default queuing values will be programmed to 1% bandwidth on the 10G interface. When stream reservation happens via MSRP, the ports will be moved accordingly from the boundary to the core port and the calculated bandwidth will be reserved for the outgoing interfaces for given streams. If a port is enabled with a capture feature like SPAN, RSPAN or Wireshark, there is no MSRP stream reservation. Queuing is programmed with default values of 1% for Class A & Class B of AVB traffic. Hence, all AVB traffic is rate limited to 1% of the bandwidth. |
Functions of MSRP
MSRP performs the following functions:
-
Allows Talkers to advertise Streams and Listeners to discover and register for Streams.
-
Establishes a path through an Ethernet between a Talker and one or more Listeners.
-
Provides guaranteed bandwidth for AVB Streams.
-
Guarantees an upper bound on latency.
-
Discovers and reports the worst case end-to-end latency between the Talker and each of its Listeners.
-
Reports failure reason and location when a path between the Talker and a Listener cannot satisfy bandwidth requirements.
-
Supports multiple classes of traffic with different latency targets.
-
Protects best effort traffic from starvation by limiting AVB traffic.
-
MSRP Talker declarations are not forwarded along the STP blocked ports.
-
MSRP listens to the STP TCN notification to generate MSRP declarations tear /modify / establish streams.
Information about QoS HQoS
AVB networks guarantee bandwidth and minimum bounded latency for the time-sensitive audio and video streams. AVB defines Class A and Class B as the time-sensitive streams, based on the worst-case latency targets of the traffic from talker to listener.
-
SR-Class A: 2ms
-
SR-Class B: 50ms
The sum of the worst-case latency contributions per hop should result in an overall end-to-end latency of 2 ms or less for SR-Class A and 50ms or less for SR-Class B. A typical AVB deployment of 7 hops from talker to listener meets these latency requirements.
The priority code points map the traffic to the specific stream. Frame forwarding behavior is based on this priority. A credit-based shaper is used to shape the transmission of these streams in accordance with the bandwidth that has been reserved on a given outbound queue so that the latency targets are met.
-
Allow a parent class to shape multiple queues in a child policy
-
Apply specific policy map actions on the aggregate traffic
-
Apply class-specific policy map actions
Note |
You should not modify the PCP in child policy to map with PCP configured in Parent Policy, e.g. SR Class A cos 3 and SR Class B Cos 2. |
Hierarchical Policing
Hierarchical policing is supported on ingress and egress interfaces. Hierarchical QoS separates the SR and Non-SR class related rules into parent and child policies respectively. AVB SR classes are completely controlled by MSRP client and hence, parent policies containing SR class attributes are governed by MSRP. The end user has complete control over child policies which contain Non-SR class attributes and can modify only the child policies.
AVB HQoS child policies are user modifiable and NVGENed to preserve the configuration if user saves the configuration to the startup-config. So, AVB HQoS child policy configurations are retained even after reload.
Information about Multiple VLAN Registration Protocol (MVRP)
Multiple VLAN Registration Protocol (MVRP) is an application based on MRP. MVRP provides a mechanism for dynamic maintenance of the contents of Dynamic VLAN Registration Entries for each vlan ids, and for propagating the information they contain to other Bridges. This information allows MVRP-aware devices to dynamically establish and update their knowledge of the set of vlan ids associated with VLANs that currently have active members, and through which Ports those members can be reached.
MVRP, from an AVB perspective, is mandatory on the talkers and the listeners. Independent of AVB, MVRP is an IEEE 802.1Q requirement on the VLAN-aware switches. However, manual configuration of VLANS on the switches is sufficient for AVB.
Note |
VTP should be in the disabled mode or transparent mode for MVRP to work. |