In the IT world, redundancy has been well understood for decades: dual uplinks, LACP bundles, stacking, spanning tree, and failover mechanisms that work seamlessly for TCP‑based traffic. For traditional data networks, these methods work perfectly.
But when this same mindset is applied to AV‑over‑IP, misunderstandings and unexpected technical problems arise.
A recent real‑world example illustrates this perfectly: a reseller believed they had created a redundant AV network by placing one switch at front‑of‑house, another on stage, and connecting the two switches with a pair of fiber links.
The assumption: “Two links means redundancy.”
The reality: this is not how redundancy works in AV‑over‑IP—especially not for realtime UDP‑ and RTP‑based audio and video flows, where clocking, timing, and uninterrupted streaming are critical.
Let’s break down the misconception and clarify what does and does not work.
1. The Biggest Misconception: Sticking to TCP Thinking
Most IT networks operate around the behavior of TCP/IP, which provides:
- retransmissions
- acknowledgements
- flow and congestion control
- packet‑order correction
Thanks to these mechanisms, TCP‑based applications can tolerate short disruptions, reconvergence events, or link failovers.
AV‑over‑IP does not use TCP. Instead, AV traffic relies on:
- UDP streams
- RTP‑based audio and video frames
- PTPv2 clocking (IEEE 1588)
These protocols:
- never retransmit
- tolerate almost no jitter
- cannot handle micro‑interruptions
- cannot absorb topology changes or STP reconvergence
- fail instantly with audible and visible dropouts
Where TCP can “wait,” AV traffic must flow continuously and with exact timing.
2. Why Traditional IT Redundancy Techniques Do Not Work for AV
Many engineers design AV‑over‑IP networks the same way they design IT networks, using:
- stacking / virtual chassis
- LACP bundles
- RSTP / MSTP
These technologies are perfectly suitable for TCP data traffic, but they introduce issues in realtime AV workflows.
2.1 A Note on Stacking
Stacking is often perceived as a redundancy model, but in AV‑over‑IP it is not suitable.
Key reasons:
- Stacked members share control‑plane functions, which destabilizes PTPv2 clocking and prevents it from operating correctly.
- Control‑plane latencies in a stack are too high for time‑critical AV streams.
- A stack behaves as one single system, introducing multiple single points of failure.
For these reasons, stacking is not considered reliable redundancy in AV networks.
2.2 LACP in AV Networks: Useful, but Not a Redundancy Mechanism
LACP is widely used across IT and AV networks to bundle bandwidth.
In NETGEAR AV switches, LACP is extremely stable and highly effective for solving uplink bottlenecks and scaling throughput.
But for AV‑over‑IP redundancy, the LACP protocol falls short.
Why?
- LACP balances per flow → AV streams are almost always single‑flow, meaning they still traverse only one physical link.
- During link failover, micro‑interruptions occur, because LACP must remove the failed link from the bundle.
→ This causes RTP clock information to drop and AV streams to disappear. - LACP therefore provides excellent scalable bandwidth, but not realtime‑safe redundancy.
In summary:
Use LACP for throughput—not for failover.
2.3 RSTP and MSTP Disrupt Realtime Traffic
- Convergence time is far too slow for AV workloads.
- State transitions introduce packet loss.
- PTP domains destabilize during any STP topology change.
RSTP/MSTP should not be part of an AV‑over‑IP core.
3. Redundancy Does Not Start With Dual Switches — It Starts With Power
Many AV issues are not network problems at all—they’re power problems.
3.1 Step One: Redundant Powering
- dual PSUs
- independent electrical circuits/phases
- UPS where appropriate
A redundant uplink does nothing if the switch itself loses power.
3.2 Step Two: AV‑Specific Network Design
- correct QoS configuration
- guaranteed per‑stream bandwidth
- stable IGMP architecture
- multicast‑capable switch fabric
- properly designed PTP domains
3.3 Step Three: Only Then Think About Network‑Level Redundancy
Such as:
- A/B networks
- separated physical paths
- mirrored switching fabrics
- devices with dual‑NIC redundancy (where available)
Redundancy is an architectural strategy, not a checklist item.
4. The Harsh Reality: Many AV Devices Have Only One Network Port
Many AV devices still ship with:
- a single NIC
- no Audio Redundancy Mode
- no SMPTE ST 2022‑7 support
Meaning:
no device‑level redundancy — a design constraint that must be accounted for.
5. So What Does Work? AV‑Native Redundancy Principles
5.1 No Single Point of Failure in the Core
- separate A and B networks and/or MLAG
- fully isolated power paths
- physically separated fiber routes
5.2 Two Fully Independent Signal Paths
Borrowed from broadcast standards:
- A = primary
- B = secondary
- Devices autonomously switch
- Failover is seamless
- Supported by Dante, AES67 with ST 2022‑7, SMPTE ST 2110
5.3 Test Redundancy the Same Way You Test Audio
- pull a cable (live)
- reboot a switch
- observe PTP behavior
- watch IGMP flow changes
- measure jitter and latency
If the stream doesn’t drop, your redundancy works.
Conclusion
True redundancy in AV‑over‑IP requires a fundamentally different approach than traditional IT:
- AV uses UDP/RTP, not TCP
- realtime traffic tolerates zero interruptions
- Stacking, LACP, and STP are not suitable failover mechanisms
- redundancy begins with power and design, not cabling
- many AV devices support only a single network path
- the most robust design remains a fully separated A/B network and/or MLAG architecture
The bottom line:
Only by understanding realtime AV protocols can you design a network that delivers true redundancy.
Eric Lindeman, NETGEAR ProAV Staff Systems Engineer Benelux
For more information about NETGEAR AV Switching, please contact the NETGEAR Pro AV Design Team via email: ProAVdesign@netgear.com
If you’d like to delve deeper into AV over IP switching, I invite you to check out our Online Academy via the link: https://academy.netgear.com/
On our training portal, you can find both AV and IT-related training courses. These courses are free to attend after registration, and at the end of each course, you can take an exam to earn a certificate.
