Blocking Player when Server not Saturated

• Previous message: Jason D: "Windows Media SNMP Objects"
• In reply to: Ravi Raman: "Blocking Player when Server not Saturated"
• Messages sorted by: [ date ] [ thread ]

Date: Wed, 6 Oct 2004 11:50:21 -0700

Ravi,

I tried backing off the Load Simulator streams more, as
you suggested. It turns out that even with NO Load
Simulator streams, this Dust_to_Glory720.wmv VBR clip
won't play from the server without stopping to rebuffer!
It's as if the WM9 Player isn't pacing itself correctly
and calling for more data in time to be displayed. I'm
playing a Microsoft WM9 clip from Microsoft's web page,
it's being served by WM9 Services on a server that will
serve 900Mbps to the Load Simulator, and it's being played
by a Microsoft WM9 Player over a GigE network - but the
stream keeps stopping to rebuffer. Something is wrong!

Steve

>-----Original Message-----
>Steve,
>
>VBR doesn't lend itself to streaming very well, so you
>should not expect as good a streaming experience as CBR
>content (let alone under stress conditions). Given this
>info, any number of things can go wrong and I am not sure
>what you can expect out of it.
>
>Instead of backing down from 80 to 70, have you checked
>with different number of load clients (say with 20
>clients, 40 clients, 60 clients, 65 clients etc..) to see
>how the WMP experience is? At what point does it go bad?
>
>Thx,
>Ravi
>--
>This posting is provided "AS IS" with no warranties, and
>confers no rights.
>
>
>
>
>
>
>
>
>
>>-----Original Message-----
>>Hi Ravi,
>>
>>The streams are VBR, but average 9.5Mbps.
>>
>>I'm dividing the load simulation between two clients
each
>>running Load Simulator. I have a separate GigE NIC in
>the
>>server for each of these clients and each NIC shows
less
>>than 50% utilization. And each client shows less than
>50%
>>CPU utilization while pulling 41 streams from the
server,
>>so it doesn't look like the clients or the network are a
>>bottleneck.
>>
>>What puzzles me is that after I back the load down from
>82
>>to 70 streams, I can't actually "play" one more stream
on
>>a third client, using Windows WM9 Player, without the
>>player stopping several times to buffer. According the
>>Performance Monitor readings the server was happy
>>supplying 9.5Mbps for all 82 streams, so when I reduce
>the
>>server load to 70 streams there should be plenty of
>>bandwidth available to satisfy a real Player playing a
>>71st stream without blocking. Load Simulator is telling
>me
>>everything is fine, but in the real world, it's not.
>>What's going on?
>>
>>
>>Thanks
>>Steve
>>
>>
>>
>>>-----Original Message-----
>>>I presume the content you are using is CBR (constant
bit
>>>rate) and not VBR.
>>>
>>>Are you using a single Load simulator to do your
>testing?
>>>How much is the Load simulator machine being stressed
>>>(memory and CPU wise)?
>>>
>>>I think that Late sends may happen if the client is not
>>>consuming data fast enough (not 100% sure about this)
>>>because the client machine is stressed out. In that
>>>case, you might want to spread the load simulator on
to
>>>multiple computers to test.
>>>
>>>Thx,
>>>Ravi
>>>--
>>>This posting is provided "AS IS" with no warranties,
and
>>>confers no rights.
>>>>-----Original Message-----
>>>>My WM9 server is able to deliver 82 "Dust-to-Glory"
>>>clips
>>>>(9.5Mbps each) to Load Simulator running on a client
>>>>across a GigE network. That is, the Current Player
>>>Send
>>>>Rate in Performance Monitor at the server averages 82
x
>>>>9.5Mbps = 779Mbps, and if I tell Load Simulator to
sink
>>>>additional streams the Current Player Send Rate
>>>increases
>>>>less than 9.5Mbps/stream, telling me the server is
>>>finally
>>>>starting to saturate.
>>>>
>>>> Now, if I tell Load Simulator to only pull 70
>streams,
>>>>the Current Player Send Rate at the server drops to 70
>x
>>>>9.5Mbps = 665Mbps and we should have bandwidth to
>spare.
>>>>But Load Simulator still reports Late Sends from the
>>>>server. I also notice that if I try to play an
>>>>additional Dust-to-Glory stream on a WM9 Player at
>>>another
>>>>client on the network while Load Simulator continues
>>>>sinking the 70 other streams from the server, the
>>>playing
>>>>stream blocks periodically at the Player: it can't get
>>>the
>>>>data it needs, even though the server should have
>>>>bandwidth available for 82-70 = 12 more clips.
>>>>
>>>>My question is, why am I getting these Late Sends and
>>>why
>>>>is the playing clip blocking when I should have plenty
>>>of
>>>>horsepower available at the server? What is the rule
>of
>>>>thumb to predict how many clips you can play without
>>>>blocking?
>>>>
>>>>Thanks for any thoughts.
>>>>
>>>>Steve
>>>>
>>>>.
>>>>
>>>.
>>>
>>.
>>
>.
>

• Previous message: Jason D: "Windows Media SNMP Objects"
• In reply to: Ravi Raman: "Blocking Player when Server not Saturated"
• Messages sorted by: [ date ] [ thread ]