Quantitative analysis
Let’s consider a 30-minute long movie encoded at 800kbps, we will estimate the broadband necessary to provide this stream to 50 concurrent players, using a CDN solution then RayV’s technology.
Upload requests with a classical a CDN
the formula is quite straightforward:
let:
- d be the duration of the movie (in seconds)
- Nv the number of concurrent players playing this stream
- MR the encoded movie rate (in kbps)
the total consumed upload is:
Applied to our example, we get:
(where factor 1/8000 converts kb into MB)
Upload requests with with RayV Cloud TV
The upload cost supported by the RayV grid can be split into two:
- on the one hand, the cost ofr seeding helping peers with partial chunks of the movie. We’ll call it UL[seeding]
- on the other hand the cost of direct players’ requests to edge nodes. We’ll call it UL[players]
We consider a scenario where we have enough helping peers available to provide the whole data to players (condition of completion). The player will request data from edge node only during a build-up phase while they populate their donors list.
Under these assumptions, we can determine UL[seeding] and UL[players] as follows:
- let ur be the average capacity of helping peers
- and 1/Sz the fraction of data received by each helping peer at seeding stage.
To satisfy the condition of completion, the minimal number of helping peers (Nh) required is:
As each helping peers receives 1/Sz of the movie at the seeding stage, we can derive:
On the other hand, considering the condition of completion, let’s т be the build- up time requested by a peer to establish a connection with the Sz helpingpeers requested to be fully autonomous from edge nodes.
We have:
Together, the upload formula using RayV’s Cloud OTT under the completion assumption is:
For the same example as above (50 viewers watching a 30 mins long movie at 800kbps), let’s consider that:
- we seed each helping peer with 1/Sz = 1/50th of the data;
- that the build-up time for a player to establish connections with helping peers is т = 30 seconds;
- and that each helping peer has an upload capacity of ur = 500 kbps.
Then the minimal total upload requested to serve these 50 (concurrent) players using RayV’s Cloud OTT is :
Compared to the upload required by a classical CDN (9000MB), it represents savings of 95% of the bandwidth!
And how does it look like?
This model has been successfully tested and validated on our grid for movie rates ranging from 300kbps to 2200kbps.
As an illustration, this is how players download is split between edge nodes and helping peers along an experiment where 150 helping peers are seeded, then 50 players join the stream at the time:
Stay Tuned for our next post where we’ll apply these results to a quite hot case study!
