Download v Streaming Speed

When I use http://www.bbc.co.uk/iplayer/diagnostics I get four different results, one download speed and three different streaming speeds (last test these were 10.4, 4.8, 7.8, 5.6meg). What's the difference between streaming speed and download speed and is there any significance to having four variable results

I assume that streaming has a higher priority/QoS.

Streaming doesn't need high throughput - rather just enough - but you do need it to be consistent. Downloading benefits from high speed but can tolerate pauses or variability.

You get variable results even on providers that do not do traffic management.
I believe its the difference between UDP, TCP and then using different ports, e.g. TCP 80 (HTTP) versus TCP 1729 (RTSP).

I have run it several times and the download speed is close to my normal max at 38xxx. The streaming come in at around 10500, 22500 and 22300. Then when the results come up it gives a streaming speed that is an average of the three. There must be different protocols being used as the first is always low compared to the others.

I think the streaming tests may be fixed-rate (capped).

Nomatter whether I test on a 60, 100, or 1000mb connection, they always come back as roughly 10/20/20

e.g. 60mb connection (DLM banded BT Infinity):

Download Speed: 55370 kbps
Streaming Speed (1): 10566 kbps
Streaming Speed (2): 22334 kbps
Streaming Speed (3): 20203 kbps

100mb connection (OVH hosted server):
Download Speed: 95184 kbps
Streaming Speed (1): 10938 kbps
Streaming Speed (2): 22707 kbps
Streaming Speed (3): 22246 kbps

1000mb connection (JANET GbE):
Download Speed: 352669 kbps
Streaming Speed (1): 10834 kbps
Streaming Speed (2): 22341 kbps
Streaming Speed (3): 21055 kbps

The tests also use different sources of stream. BBC in-house x 2 different types of data and then 2 CDNs.

Matt

When ripping the streams with getiplayer, an SD stream maxes out at about 20Mb and a HD stream at about 10Mb.

It seems test 1 is HD streaming speed and test 2 and 3 is SD streaming speed.