WizPnP measured streaming throughput.

PC, Mac and Linux Networking Hints and Tips.

Moderators: Gully, peteru

Post Reply
prl
Wizard God
Posts: 32705
Joined: Tue Sep 04, 2007 13:49
Location: Canberra; Black Mountain Tower transmitters

WizPnP measured streaming throughput.

Post by prl » Sun Jan 04, 2009 10:04

The ability of the Beyonwiz to act as a WizPnP streaming server for other Beyonwiz devices is fairly constrained, especially for streaming HD recordings.

In the past I've tried to characterise the streaming behaviour by streaming HD recordings from a DP-S1 to a DP-H1, and noting whether the (short) trials were successful or not.

This time, I've measured the streaming bitrate by copying a recording from the DP-S1 to a Mac, via WizPnP using getWizPnP. getWizPnP simply tries to read the WizPnP stream as fast as possible. I don't know how well this represents the streaming done when streaming to a Beyonwiz WizPnP client, but the streaming probably can't go any faster.

While this method doesn't give a direct indication of whether streaming will work or not, it's more flexible, since if you know the bitrate of any media file, you should be able to compare it with the bitrates in the table.

Bitrates and server load for WizPnP data transfer

Code: Select all

+--------------------------------------------------------------------------+
|                WizPnP average throughput Mib/s (01.05.280)               |
+-------+----------------+------------------------+------------------------+
|       |     In Menu    |       Watch HD         | Watch HD, record 1 HD  |
+-------+------+---------+------+---------+-------+------+---------+-------+
| Video |  No  |   TS    |  No  |   TS    | Watch |  No  |   TS    | Watch |
| out   |  TS  | enabled |  TS  | enabled | TS    |  TS  | enabled |  TS   |
+-------+------+---------+------+---------+-------+------+---------+-------+
|       |      |         |      |         |       |      |         |       |
|  720p | 24.8 |  24.0   | 22.4 |  18.4   | 12.8  | 16.0 |  14.4   |  9.6  |
|       |      |         |      |         |       |      |         |       |
|-------+------+---------+------+---------+-------+------+---------+-------+
|       |      |         |      |         |       |      |         |       |
| 1080i | 24.8 |  24.0   | 21.6 |  18.4   | 12.8  | 16.8 |  13.6   |  8.8  |
|       |      |         |      |         |       |      |         |       |
+-------+------+---------+------+---------+-------+------+---------+-------+
Video out is the Video A/V resolution, otherwise always HDMI/PAL, 2ch decoding.
In Menu means that the Beyonwiz is displaying the Setup menu.
Watch HD, means that the TV is decoding the service to the screen. For these tests, Prime HD (1440x1080p, 13.6MiB/s).
Watch HD, record 1 HD means that in addition to decoding TV, another HD service is being recorded. For these tests SC10 HD (1920x1080i)
No TS means timeshifting is disabled.
TS enabled, means that timeshifting is enabled, but Live TV is being watched.
Watch TS means that the TV service is being decoded from the timeshift buffer.

Other server loads.

The table is not comprehensive of all server loads. I have assumed that streaming two simultaneous HD recordings will never be feasible. Two SD recordings may be under some circumstances (needs about 12-13Mib/s total throughput).

Chase-play with only one recording running should be similar to Watch HD/Watch TS.
Viewing a recording while recording another service should be similar to Watch HD/Watch TS (TS is temporarily disabled by activity).
Chase-play of one recording while recording two programs should be similar to Watch HD, record 1 HD/Watch TS (TS is temporarily disabled disabled by activity).
Viewing a recording with no other activity should fall between Watch HD/No TS and Watch HD/Enable TS (TS is temporarily disabled disabled by activity).

Typical HD data rates.

Code: Select all

+----------------+-------------------+-------+
| Typical HD service bitrates MiB/s (Nov 08) |
|          (Canberra Black Mountain)         |
+----------------+-------------------+-------+
|     Service    |    Resolution     | Mib/s |
+----------------+-------------------+-------+
|     SBS HD     |     1280x720p     |   8   |
|     ABC HD     |     1280x720p     |  10   |
|     WIN TV HD  |    1440x1080i     |  11   |
|     SC 10 HD   |    1920x1080i     |  14   |
|     PRIME HD   |    1440x1080i     |  14   |
+----------------+-------------------+-------+
If the service bitrate is less than the WizPnP data transfer rate for a particular server load by a reasonable margin, it should be possible to stream recordings of that service with that server load.

Similarly, it should be possible to tell whether other media types can be streamed under a given server load. I'd be very interested in hearing whether the table predicts the actual streaming of video well, or not.

Technical notes
  • The data transfer rates were gathered by the (yet to be distributed) getWizPnP 0.3.3, on an iMac G5 2GHz, 1.5GB memory, copying in Beyonwiz recording format to a Firewire connected WD MyBook Home Edition 500GB HDD.
  • The data rates are Mib/s, mebibits/sec. 1 Mi = 2^20 = 1024*1024 = 1048576. 1Mi ~= 1.05M. Take care when comparing these bit rates with Mb/s 1Mib/s ~= 1.05Mb/s.
  • The actual data precision is +-0.8Mib/s, converted from +-0.1MiB/s.
  • The recording transferred was 5 minutes of ABC HD, 370MiB. The video format is unimportant, because it was simply being copied, not decoded.
  • The numbers quoted are averages for the whole transfer. The actual transfer rate at any instant varies about the average slightly. The important figure for transfer rate is the minimum rate after averaging over the streaming buffer size on the WizPnP client, but its size is unknown.
  • The transfer rate at the start of the transfer is often much lower than the average. It's uncertain whether this is a genuine phenomenon, or a bug in getWizPnP's sliding average calculation. The overall average transfer rate is not affected by any miscalculation in the sliding average.
  • The time shift lag for Watch TS was short and not carefully controlled.
  • Posts of transfer rates for other Beyonwiz servers and for other server loads are most welcome.
  • The streaming rate changes between when the Beyonwiz is outputting 720p or 1080i are very small.
Last edited by prl on Sun Jan 04, 2009 13:41, edited 1 time in total.
Peter
T4 HDMI
U4, T4, T3, T2, V2 test/development machines
Sony BDV-9200W HT system
LG OLED55C9PTA 55" OLED TV

User avatar
tonymy01
Uber Wizard
Posts: 6373
Joined: Fri Jun 01, 2007 15:25
Location: Sydney, Australia DP-S1-1TB, DP-P2-2TB, DP-T4-2TB, DP-T4-BB... too many!
Contact:

Post by tonymy01 » Sun Jan 04, 2009 10:34

Nice set of tests. Just goes to show that recording something and watching something drags the bitrates to below what Sydney HD is broadcast at (some stations are getting close to 15megabit/s for their HD!)
Regards
Tony

IanSav
Uber Wizard
Posts: 16846
Joined: Tue May 29, 2007 15:00
Location: Melbourne, Australia

Post by IanSav » Sun Jan 04, 2009 10:43

Hi Peter,

Nice research. I wonder if this can be used to somehow get Beyonwiz to improve their priority models?

Regards,
Ian.

User avatar
Gully
Moderator
Posts: 7736
Joined: Thu Aug 30, 2007 22:08
Location: Melbourne

Post by Gully » Sun Jan 04, 2009 10:58

IanSav wrote:Nice research. I wonder if this can be used to somehow get Beyonwiz to improve their priority models?
I wonder if that is wise though. After all, you don't really want a recording or watching to suffer due to streaming.

The figures for watching the timeshift buffer intrigue me - why would it give such a performance hit?
Cheers
Gully
_____________
Beyonwiz U4
Logitech Harmony Elite
Google Pixel 6 Pro

IanSav
Uber Wizard
Posts: 16846
Joined: Tue May 29, 2007 15:00
Location: Melbourne, Australia

Post by IanSav » Sun Jan 04, 2009 11:14

Hi Gully,
Gully wrote:I wonder if that is wise though. After all, you don't really want a recording or watching to suffer due to streaming.
I agree. Though there is a desire/need to get more performance from the streaming side of the Beyonwiz.

Regards,
Ian.

prl
Wizard God
Posts: 32705
Joined: Tue Sep 04, 2007 13:49
Location: Canberra; Black Mountain Tower transmitters

Post by prl » Sun Jan 04, 2009 12:08

Gully wrote:...
The figures for watching the timeshift buffer intrigue me - why would it give such a performance hit?
The networking throughput, as far as we know, is not being limited by coming up against hardware limits, but artificially limited so the local functions don't come up against hardware limits. Perhaps the settings are a bit too conservative.
Peter
T4 HDMI
U4, T4, T3, T2, V2 test/development machines
Sony BDV-9200W HT system
LG OLED55C9PTA 55" OLED TV

prl
Wizard God
Posts: 32705
Joined: Tue Sep 04, 2007 13:49
Location: Canberra; Black Mountain Tower transmitters

Post by prl » Sun Jan 04, 2009 12:12

tonymy01 wrote:Nice set of tests. Just goes to show that recording something and watching something drags the bitrates to below what Sydney HD is broadcast at (some stations are getting close to 15megabit/s for their HD!)
Regards
Remember that 15Mb/s is 14.2Mib/s, which is close to what I've measured for the better regional HD services.

It's interesting to note that the WIN HD bitrate for 1080i is only slightly better than the ABC's bitrate for 720p.

I wonder when the bitrate squeeze for the extra Freeview channels on the commercials will come?
Peter
T4 HDMI
U4, T4, T3, T2, V2 test/development machines
Sony BDV-9200W HT system
LG OLED55C9PTA 55" OLED TV

IanSav
Uber Wizard
Posts: 16846
Joined: Tue May 29, 2007 15:00
Location: Melbourne, Australia

Post by IanSav » Sun Jan 04, 2009 12:25

Hi Peter,
prl wrote:I wonder when the bitrate squeeze for the extra Freeview channels on the commercials will come?
Probably when those channels start transmitting. ;)

Regards,
Ian.

Post Reply

Return to “Networking your Beyonwiz DP Series”