Leaf & spine networks have started taking over data centres in the last few years. It’s no secret that people prefer scale-out over scale-up solutions and you can see a similar approach in ST 2110 networks, when large monolithic video switches are replaced with smaller leaf and spine switches.
Leaf and spine refers to networks where a number of main, high throughput switches link to a number of smaller switches. These smaller switches tend to be aggregators and offer the promise of cheaper ports delivered closer to your equipment. The alternative to leaf & spine is monolithic switches which do have their merits, but are certainly not always the right choice.
To provide non-blocking switching in leaf & spine networks you need an SDN controller that orchestrates media flows. Advances in SDN capabilities have led to the emergence of “Purple” network architectures. In this video Gerard Phillips from Arista shows how it differs from a “Red/Blue” architecture, how path diversity is maintained and how ST 2110 IP live production or playout applications could benefit from it.
It’s important to be aware of the different uses of Layer 2 vs Layer 3:
• Layer 2 devices are typically used for audio networks like Dante and RAVENNA. A layer 2 network is a simple, scalable and affordable choice for audio flows where there are no challenges in terms of bandwidth. However, this type of network doesn’t really work for high bit rate live production video multicast since all multicasts need to be delivered to the IGMP querier which isn’t scalable.
• Layer 3 have distributed IGMP management since PIM is used on each router to route multicast traffic, so there is no more flooding network with unnecessary traffic. This type of network works well with high bit rate video multicasts, but as IGMP is not bandwidth aware, it’s best to use an SDN system for flow orchestration.
Gerard then looks at resilience:
Using 2022-7 seamless switching (plus a robust monitoring system that can provide quick, accurate information to resolve the issue)
Providing redundancy (redundant PSU, fans, fabric modules etc., redundant links between switches, ensuring that routing protocol or SDN can use these “spares”)
Dividing up failure domains
Using leaf and spine architecture (routing around failed components with SDN)
Using resilient IP protocols (BGP, ECMP)
The talk finishes up discussing the pros and cons of the different architectures available:
Monolithic systems which are non-blocking, but have a wide failure domain
Monolithic – expansion toward spine and leaf with SDN for non-blocking switching
Leaf & spine with air-gapped Red and Blue networks
Leaf & spine hybrid with Purple switches connected to both Red and Blue spines to support single homed devices
Leaf & spine Purple. Here, red and blue flows are connected to physically separate switches, but the switches are not identified as red and blue anymore. This is a converged network and an SDN controller is required to provide diverse paths flows to go to two different spines.
The transition from point-to-point SDI based infrastructure to IP essence flows requires a very different approach to fault-finding. Although new IP diagnostic tools are already available on the market, engineers need combined broadcast and IT knowledge to fully understand the flow of video, audio and data across the switching fabric – including packet jitter, latency, and buffer over/underflows causing dropped packets.
In this video Michael Waidson from Tektronix presents methodologies involved in monitoring IP media networks. The following topics are covered:
Strategies for choosing IP Address, Port Number and Payload Type for easier identification of the streams
Troubleshooting basics (fibres and SFPs types, checking switch ports)
PTP and uncompressed video go hand in hand so this primer on ST 2022 and ST 2110 followed by a PTP deep dive is a great way to gain your footing in the uncompressed world.
In the longest video yet on The Broadcast Knowledge, Steve Holmes on behalf of Tektronix delivers two talks and a practical demo for the SMPTE San Francisco section where he introduces the reasons for and solutions to uncompressed video and goes through the key standards and technologies from ST 2022, those being -6 video and -7 seamless switching plus the major parts of ST 2110, those being timing, video, audio and metadata.
After that, at the 47 minute mark, Steve introduces the need for PTP by reference to black and burst, and goes on to explain how SMPTE’s ST2059 brings PTP into the broadcast domain and helps us synchronise uncompressed essences. He covered how PTP actually works, boundary clocks, Grandmaster/Master/Slave clocks and everything else you need to understand the system,
This video finishes with plenty of questions plus a look at the GUI of measurement equipment showing PTP in real life.
With the SMPTE 2110 suite of standards largely published and the related AMWA IS-04 and -05 specifications stable, people’s minds are turning to how to implement all these standards bringing them together into a complete working system.
The JT-NM TR-1001-1 is a technical recommendation document which describes a way of documenting how the system will work – for instance how do new devices on the network start up? How do they know what PTP domain is in use on the network?
John Mailhot starts by giving an overview of the standards and documents available, showing which ones are published and which are still in progress. He then looks at each of them in turn to summarise its use on the network and how it fits in to the system as a whole.
Once the groundwork is laid, we see how the JT-NM working group have looked at 5 major behaviours and what they have recommended for making them work in a scalable way. These cover things like DNS discovery, automated multicast address allocation and other considerations.