P2P Low Latency AV streaming

[QuoVadis]

In short: I am recommending LOLA (LOw LAtency audio visual streaming system) to those who wish to set up a peer-to-peer A/V live stream with low latency (few miliseconds). Initially developed by Conservatorio di Musica Giuseppe Tartini from Trieste (Italy) in collaboration with GARR, LOLA can be used for free for educational purposes (license can be requested from educational establishments) and for a fee for commercial purposes. It is an excellent, high-end live stream platform that enables two peers to have (probably) the best quality live stream possible in 2015. Potential uses: musical collaboration (initially developed for this very purpose), high-end communications, medical, etc.

Description: The software is quite straight forward to use for any beginners and further information with screenshots and guidance can be found in their comprehensive and very useful manual developed over the years. However, the hardware side of things is slightly more complicated and more expensive:

- The workstation being used needs to be a powerful one. Further information can be found in the manual link above but we have successfully used so far the following:

- Intel Xeon CPU E5-1650 (12 core) @ 3.20 GHz

- 16GB RAM

- 4GB NVIDIA Quadro K2000 video card

- 2 x 512GB SSD hard drive

- RME HDSP 9632 PCI-E sound card

- Bandwidth: minimum 1Gbps

The LOLA software is a bit touchy about the soundcards being used and some may create more latency than others. This was was recommended by other current users and it seems to do the job quite well. The card also comes with some particularly useful tools in setting everything up.

Also the video aspect is important, for obvious reasons. Again, following advice from previous users, we used a Ximea camera with a Tamron lense connected to the computer through a USB 3 camera cable.

Frankly it's quite impressive that such a tiny camera and lenses can deliver a whooping HD video (different resolutions) at 60fps. So far we have used 1 Ximea camera for the live stream using a 1280 x 720 resolution with 60fps which was using approximately 500 Mb/s bandwidth. This may be problematic if your internet connection is not good enough. The other important aspect is that your traffic must not be filtered or routed through firewalls or stuff to avoid any latency.

In terms of delay, for a P2P identical connection (as the one described above) over a 200km distance the delay was 9msec. For a connection over 700km distance the delay was between 11-19. However, as the traffic was passing purely through the JANET (UK) network, other networks and CDNs may impact the connection in different ways. However, this is still very much acceptable as for many people few miliseconds delay is quite negligible.

We also recommend using good quality cables as much as you can and if possible reduce the number of adapters to a minimum as each of these factors contribute to the delay:

It supports various audio inputs but regular (good quality) XLR microphones routed through an amplifier/mixing desk will do the job. As this is live sound rather than recorded, careful consideration needs to be given to the audio set-up being used as well as collaboration with the users to ensure best positioning in the room/studio/hall.

LOLA has an in-built recording system for each session. However, they video is recorded frame by frame in pictures and the sound separately in high-quality uncompressed sound files. It therefore requires quite a bit of post-production work to have a decent recording.

Conclusion: An excellent high-end live-streaming platform that brings the best possible audio and video to the table. However, pricing can be quite prohibitive and it requires both parties to have a similar set-up. At the moment it does not support multi-cam nor does it support multiple participants. That being said, it is arguably the best live streaming platform for low latency in the world. Highly recommended for researchers in these fields, musicians, medical companies, legal system and other potential scenarios where funding wouldn't be a a major limitation.