Macromedia Flash Server - Strange variance in time to build buffer

This is Interesting: Free IT Magazines  
Home > Archive > Macromedia Flash Server > August 2006 > Strange variance in time to build buffer





You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

Author Strange variance in time to build buffer
Will Law

2006-08-11, 1:12 pm


I have a simple netstream playing a 300kbps file from a FMS2 server. I
increase the setBufferTime() linearly from 1 to 10 seconds and time how long
it takes to go from netstream.play.start to netstream.buffer.full i.e how
long it takes to build the buffer. Here are the results

BufferTime(s) Milliseconds-to-build
1 877
2 1945
3 952
4 888
5 928
6 1910
7 1952
8 2938
9 1937
10 2953

This is not what I expected at all. I would expect (given that my connection
to the server is stable, the server is on a different network and it is
unloaded except for my test) that it would take twice as long to build a 2s
buffer vs. a 1s buffer and 10x as long to build a 10s buffer as a 1s. But
the results are all over the place. Can you explain this ?

Cheers

Will

________________________________________
_______
FlashComm-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org
To change your subscription options or search the archive:
http://chattyfig.figleaf.com/mailman/listinfo/flashcomm

Brought to you by Fig Leaf Software
Premier Authorized Adobe Consulting and Training
http://www.figleaf.com
http://training.figleaf.com

Stefan Richter

2006-08-11, 1:12 pm

I'd agree to a point here. Yes the buffer for say 5 seconds should take
longer to fill than for 2 seconds (in your example it doesn't and that is
weird).
However given the fact that - if I understand this correctly - a 1 second
buffer holds enough data to play 1 second of video (but doesn't necessarily
take 1 second to fill) and that bitrates may fluctuate (VBR encoding) then
it would make sense that it's not as simple as saying: 5 seconds of buffer
take 5 times as long as 1 second.

Having said that (and we talked about this today), bandwidth tests against a
local or remote FMS box return greatly varying results for me (and I'm not
alone). This in mind it would not surprise me if the server somehow spits
out data at different rates, maybe depending on load or whatever.
I'm speculating here though...

What happens when you run your tests over and over using the same server
state and video? Do you consistently get weird numbers like this?

Did you use a CBR encoded video or a VBR?

Stefan




> -----Original Message-----
> From: flashcomm-bounces-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org
> [mailto:flashcomm-bounces-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org] On Behalf Of Will Law
> Sent: 11 August 2006 17:11
> To: 'FlashComm Mailing List'
> Subject: [FlashComm] Strange variance in time to build buffer
>
>
> I have a simple netstream playing a 300kbps file from a FMS2
> server. I increase the setBufferTime() linearly from 1 to 10
> seconds and time how long it takes to go from
> netstream.play.start to netstream.buffer.full i.e how long it
> takes to build the buffer. Here are the results
>
> BufferTime(s) Milliseconds-to-build
> 1 877
> 2 1945
> 3 952
> 4 888
> 5 928
> 6 1910
> 7 1952
> 8 2938
> 9 1937
> 10 2953
>
> This is not what I expected at all. I would expect (given
> that my connection to the server is stable, the server is on
> a different network and it is unloaded except for my test)
> that it would take twice as long to build a 2s buffer vs. a
> 1s buffer and 10x as long to build a 10s buffer as a 1s. But
> the results are all over the place. Can you explain this ?
>
> Cheers
>
> Will
>
> ________________________________________
_______
> FlashComm-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org
> To change your subscription options or search the archive:
> http://chattyfig.figleaf.com/mailman/listinfo/flashcomm
>
> Brought to you by Fig Leaf Software
> Premier Authorized Adobe Consulting and Training
> http://www.figleaf.com http://training.figleaf.com
>



________________________________________
_______
FlashComm-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org
To change your subscription options or search the archive:
http://chattyfig.figleaf.com/mailman/listinfo/flashcomm

Brought to you by Fig Leaf Software
Premier Authorized Adobe Consulting and Training
http://www.figleaf.com
http://training.figleaf.com

Troy Gardner

2006-08-12, 7:11 pm

Part of what you may be seeing is that data is sent in chunks, even thoug=
h the
stream plays at a continous rate. e.g. it may fill in 100bit jumps...paus=
e for
a second then get 300bits. wait 2 seconds get 1500 bits.=20

I've noticed lots of varience on the bufferTime with what's in the
bufferLength, it's actually kinda frustrating to deal with. Sometimes it'=
s
consistently many seconds over the buffer sometimes it's consistenlty bel=
ow it.
I tried the double buffer length strategy as suggested by Adobe, and foun=
d it
didn't increase the buffer size until it had emptied the current buffer(!=
).

I've asked this before, but didn't get a response. How are people doing l=
ow
bandwidth (e.g. dialup user testing)?

Troy Gardner -"How you live your seconds, is how you live your days, is h=
ow you live your life..."

http://www.troygardner.com -my world
http://www.troyworks.com - building Rich Internet Applications
http://www.intrio.com -helping bridge the gap between the humans and mach=
ines. Home of the Flickey=99
________________________________________
_______
FlashComm-1Ss2GqJETD3yZ38Mhd3e/9ZfFG6BLHNm@public.gmane.org
To change your subscription options or search the archive:
http://chattyfig.figleaf.com/mailman/listinfo/flashcomm

Brought to you by Fig Leaf Software
Premier Authorized Adobe Consulting and Training
http://www.figleaf.com
http://training.figleaf.com

Sponsored Links






Free braindumps | Software forum | Database administration forum

Copyright 2003 - 2008 webservertalk.com