LiveIP & Cloud Production – 25Gb and Beyond

The “Cisco Global Cloud Index 2015 – 2020” estimates that Hyperscale datacentres will account for 53% of total global network traffic by 2020. These datacentres are demanding ever higher network bandwidth.

That’s great news for Media.

The requirements of my most recent engagement were so demanding that the only way to realise the target architecture was to leverage products and technologies which were in development and, where product didn’t exist, initiate R&D work to fill the gaps.

The facility was to be an all IP “LiveIP” production facility and, when I started in 2015, 100Gb PSM4 and Single Mode 25Gb were one roadmap area. At the time, no one was talking about building 25Gb capable endpoints. Indeed, IEEE 802.3cc has only recently been approved.

When I started the journey, I’d expected most endpoints to be 10Gb attached, with some on 40Gb and 100Gb. By the time we were done with the design ~70% of endpoints were PSM4 / 25Gb attached.

What happened?

Early on, one vendor sounded me out on 25Gb endpoints. My response was an emphatic, “Yes, I want this”. This meant a lower cost, lower power design, with greater density and significantly more efficient use of structured cabling. 25Gb is also a good fit for UHD.

When presented with the option, engineers chose to deploy 25Gb en masse over 10Gb and 40Gb. There’s a lot of pent up demand for higher bandwidth to support Media use cases but do you really want to flood wire your datacentre with MPO?  When 25Gb capable FPGAs start to arrive at a price point similar to their 10Gb counterparts, 25Gb will be driven further into the mainstream.

So, what’s next?

Hyperscale datacentres are driving technology and standards development towards ever greater bandwidth. In December 2017, the IEEE 802.3bs standard was approved. IEEE 802.3bs will deliver 200Gb / 400Gb over an 8 fibre, PSM4 interface (100Gb over a pair of fibres – see also IEEE 802.3cd/100GBASE-DR).

Bandwidth intensive UHD workflows and the desire to build cloud / layered infrastructure provide drivers for 100/400Gb adoption in media production.

IEEE 802.3bs promises to provide the basis for multi-terabit switch uplink bandwidth, which is necessary for Clos and stretched network fabric architectures to become a more attractive choice for uncompressed live production use cases. It will be an important enabler for further network convergence and cloud based production architectures.

Where MPO is fine for switch to switch uplinks, it’s less suitable for connecting to endpoints. Delivery of 100Gb over two fibres (IEEE 802.3cd) reduces the need to take MPO out to endpoints, simplifying structured cabling design, reducing costs and resulting in greater reliability.

Those conceiving media datacentres today should look towards deploying infrastructure to support IEEE 802.3bs and IEEE 802.3cd, as well as those manufacturers with these standards on their roadmap.