asterisk users - May 2005 - Ast 1.0.7, IP-500's with unmanaged switch...remote end missing bits of audio

I thought I would throw this out there and see if anyone has any
ideas...I have the same problem at 2 locations.
 
The complaint from the users is that calls "cut out", "kinda like
when
you have spotty cell coverage". Doesn't seem to matter whether the call
is incoming or outgoing, although it might be true that my users hear
the remote party cut out, while the remote party doesn't notice the same
from my users...
 
Location 1:
 
- SDSL 1.5 Mpbs with static IP, Netopia 4652 SDSL router (enabled
"Prioritize Delay Sensitive Data" to recognize tos=lowdelay per
Netopia
support)
- Dell PowerEdge SC420 with TDM04B (currently only using one port.  the
single analog line is call forward on busy to my IAX provider)
- Asterisk CVS-v1-0-02/22/05 using IAX to connect to my provider over
the public internet - I have run pings for an extended period of time
against my provider's server and get no packet loss.
- In IAX.conf: tos=lowdelay, jitterbuffer=yes, also enabled "Prioritize
Delay Sensitive Data" on the Netopia to support tos=lowdelay per Netopia
support
- Average ping time to my provider: 160 ms with no packet loss
- 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1 using ulaw
only
- Small Business Server 2003 set up as DC for the network
- Two network laser printers
- 24 port unmanaged switch (all phones are home run back to a patch
panel, patched from there into a switch port.  The DSL modem, printers
and server are patched into the switch in the same way)
- 8 pc's running XP Pro, all plugged into the switch port on the back of
the IP-500's
 
Location 2:
 
- Full rate data T1
- Dell PowerEdge SC1420 with no TDM hardware at all (this location
connects SIP directly to the T1 providers BroadSoft switch and does not
go over the public internet)
- Asterisk 1.0.7 - using SIP to connect with my provider (not across
public internet, not natted since the Cisco IAD does the SIP mangling
for us)
- Average ping time to the broadsoft switch: 42 ms
- 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1 using ulaw
only
- Small Business Server 2003 set up as DC for the network
- One network printer
- 24 port unmanaged switch (all phones are home run back to a patch
panel, patched from there into a switch port.  The DSL modem, printers
and server are patched into the switch in the same way)
- 8 pc's running XP Pro, all plugged into the switch port on the back of
the IP-500's
 
 
As you can see, the only commonalities are Dell hardware (but not
models), Asterisk (but not versions), IP-500's (including sip and
bootrom version), SBS 2003, 24 port unmanaged switch, the fact that all
the pc's are plugged into the switch ports on the phones.
 
Same symptoms at both locations.  I cannot determine any specific causes
(ie it doesn't seem to be inbound vs. outbound, etc).  I have checked
all the pc's for viruses and worms, changed switch ports, etc...the only
theory I have right now is that since the Polycoms give priority to
outbound phone traffic vs a connected PC, that outbound voice is getting
the QOS it needs.  Versus inbound voice which gets no priority treatment
once it hits either LAN since the switch can't do any QOS.  Am I on the
right track with this theory?  Do I need to try a managed switch, giving
priority to voice to make sure that both incoming and outgoing voice
packets are preferred?  On a side note, at what point (size - number of
clients) is a managed switch recommended?  required?
 
If I'm off base with the QOS theory, what else should I be looking at?  
 
 
Marty
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.digium.com/pipermail/asterisk-users/attachments/20050503/f19a3edd/attachment.htm

Hi Marty -
> The complaint from the users is that calls "cut out", "kinda
like when
> you have spotty cell coverage". Doesn't seem to matter whether the
call
> is incoming or outgoing, although it might be true that my users hear
> the remote party cut out, while the remote party doesn't notice the 
> same
> from my users...
>
> Location 1:
>
> - SDSL 1.5 Mpbs with static IP, Netopia 4652 SDSL router (enabled
> "Prioritize Delay Sensitive Data" to recognize tos=lowdelay per
Netopia
> support)
> - Dell PowerEdge SC420 with TDM04B (currently only using one port.  the
> single analog line is call forward on busy to my IAX provider)
> - Asterisk CVS-v1-0-02/22/05 using IAX to connect to my provider over
> the public internet - I have run pings for an extended period of time
> against my provider's server and get no packet loss.
> - In IAX.conf: tos=lowdelay, jitterbuffer=yes, also enabled
"Prioritize
> Delay Sensitive Data" on the Netopia to support tos=lowdelay per 
> Netopia
> support
> - Average ping time to my provider: 160 ms with no packet loss
> - 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1 using 
> ulaw
> only
> - Small Business Server 2003 set up as DC for the network
> - Two network laser printers
> - 24 port unmanaged switch (all phones are home run back to a patch
> panel, patched from there into a switch port.  The DSL modem, printers
> and server are patched into the switch in the same way)
> - 8 pc's running XP Pro, all plugged into the switch port on the back 
> of
> the IP-500's
>
> Location 2:
>
> - Full rate data T1
> - Dell PowerEdge SC1420 with no TDM hardware at all (this location
> connects SIP directly to the T1 providers BroadSoft switch and does not
> go over the public internet)
> - Asterisk 1.0.7 - using SIP to connect with my provider (not across
> public internet, not natted since the Cisco IAD does the SIP mangling
> for us)
> - Average ping time to the broadsoft switch: 42 ms
> - 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1 using 
> ulaw
> only
> - Small Business Server 2003 set up as DC for the network
> - One network printer
> - 24 port unmanaged switch (all phones are home run back to a patch
> panel, patched from there into a switch port.  The DSL modem, printers
> and server are patched into the switch in the same way)
> - 8 pc's running XP Pro, all plugged into the switch port on the back 
> of
> the IP-500's
I was rather interested to read about the problem you're having.  We 
have a similar setup, and we've been experiencing the same problems.  
We have a Dell PE1600SC in the main office, and SC420's in our remote 
offices.  Main office has a full T1 with a Cisco 1721.  One remote 
office has a full T1 (but we're actually using a DSL backup for IAX 
traffic) with a Cisco 1751, and the other remote office has SHDSL with 
a Linksys.  There are a mixed bag of managed and unmanaged 3Com 
switches, and one office has 3Com hubs.  We are running the same CVS 
HEAD version (CVS-HEAD-04/09/05-08:01:41) on all machines.  I was able 
to mitigate things with a lot of jitterbuffer on the IAX, QoS rules on 
our router, and using our idle backup ISP connections for all the IAX 
traffic.  It is to the point where the sound quality is acceptable, but 
it is not commercial quality - not like Vonage or Broadvoice.  We still 
have drops and pops, and the quality is close to what it should be, but 
the remote IAX connections are NOT as good as the "local" calls on the
same asterisk box, even though it's all the same codec.

This makes me wonder if perhaps the SC420's are the problem.  I know 
that I had a bit of a problem getting a driver installed for the 
broadcom NIC that is in them.  I think that the linux driver is a very 
recent addition for this chipset.  Do you have any other suspicions?

I did notice that you said you have 160 ms latency between your 
offices.  That is a bit high.  Our is only 20-40 ms between our various 
offices.

Maybe I'll try a different NIC in one of the SC420's.

- Noah

> Hi Marty -
> 
> > The complaint from the users is that calls "cut out", 
> "kinda like when 
> > you have spotty cell coverage". Doesn't seem to matter
whether the
> > call is incoming or outgoing, although it might be true 
> that my users 
> > hear the remote party cut out, while the remote party 
> doesn't notice 
> > the same from my users...
> >
> > Location 1:
> >
> > - SDSL 1.5 Mpbs with static IP, Netopia 4652 SDSL router (enabled 
> > "Prioritize Delay Sensitive Data" to recognize tos=lowdelay
per
> > Netopia
> > support)
> > - Dell PowerEdge SC420 with TDM04B (currently only using one port.  
> > the single analog line is call forward on busy to my IAX provider)
> > - Asterisk CVS-v1-0-02/22/05 using IAX to connect to my 
> provider over 
> > the public internet - I have run pings for an extended 
> period of time 
> > against my provider's server and get no packet loss.
> > - In IAX.conf: tos=lowdelay, jitterbuffer=yes, also enabled 
> > "Prioritize Delay Sensitive Data" on the Netopia to support 
> > tos=lowdelay per Netopia support
> > - Average ping time to my provider: 160 ms with no packet loss
> > - 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1
using
> > ulaw only
> > - Small Business Server 2003 set up as DC for the network
> > - Two network laser printers
> > - 24 port unmanaged switch (all phones are home run back to a patch 
> > panel, patched from there into a switch port.  The DSL 
> modem, printers 
> > and server are patched into the switch in the same way)
> > - 8 pc's running XP Pro, all plugged into the switch port 
> on the back 
> > of the IP-500's
> >
> > Location 2:
> >
> > - Full rate data T1
> > - Dell PowerEdge SC1420 with no TDM hardware at all (this location 
> > connects SIP directly to the T1 providers BroadSoft switch and does 
> > not go over the public internet)
> > - Asterisk 1.0.7 - using SIP to connect with my provider 
> (not across 
> > public internet, not natted since the Cisco IAD does the 
> SIP mangling 
> > for us)
> > - Average ping time to the broadsoft switch: 42 ms
> > - 8 Polycom IP-500's running SIP 1.4.1.0040 and bootrom 2.6.1
using
> > ulaw only
> > - Small Business Server 2003 set up as DC for the network
> > - One network printer
> > - 24 port unmanaged switch (all phones are home run back to a patch 
> > panel, patched from there into a switch port.  The DSL 
> modem, printers 
> > and server are patched into the switch in the same way)
> > - 8 pc's running XP Pro, all plugged into the switch port 
> on the back 
> > of the IP-500's
> 
> I was rather interested to read about the problem you're 
> having.  We have a similar setup, and we've been experiencing 
> the same problems.  
> We have a Dell PE1600SC in the main office, and SC420's in 
> our remote offices.  Main office has a full T1 with a Cisco 
> 1721.  One remote office has a full T1 (but we're actually 
> using a DSL backup for IAX
> traffic) with a Cisco 1751, and the other remote office has 
> SHDSL with a Linksys.  There are a mixed bag of managed and 
> unmanaged 3Com switches, and one office has 3Com hubs.  We 
> are running the same CVS HEAD version 
> (CVS-HEAD-04/09/05-08:01:41) on all machines.  I was able to 
> mitigate things with a lot of jitterbuffer on the IAX, QoS 
> rules on our router, and using our idle backup ISP 
> connections for all the IAX traffic.  It is to the point 
> where the sound quality is acceptable, but it is not 
> commercial quality - not like Vonage or Broadvoice.  We still 
> have drops and pops, and the quality is close to what it 
> should be, but the remote IAX connections are NOT as good as 
> the "local" calls on the same asterisk box, even though it's 
> all the same codec.
> 
> This makes me wonder if perhaps the SC420's are the problem.  
> I know that I had a bit of a problem getting a driver 
> installed for the broadcom NIC that is in them.  I think that 
> the linux driver is a very recent addition for this chipset.  
> Do you have any other suspicions?
> 
> I did notice that you said you have 160 ms latency between 
> your offices.  That is a bit high.  Our is only 20-40 ms 
> between our various offices.
> 
> Maybe I'll try a different NIC in one of the SC420's.
> 
> - Noah
> 
Are you using vlans at any of these locations?  Just to clarify, my two
locations don't interface with each other.  They are two different
clients of ours each connecting only to the PSTN (one via IAX, the other
via SIP).  I should also mention that the "cutting-out" is not an
issue
all the time, it is intermittent and I haven't been able to pin it down
to one thing in particular (ie it has been reported with everyone in the
office and on the phones, as well as on a Sunday when only the owner was
in the office and using his phone.

Thanks,

Marty

Hi Marty -
> Are you using vlans at any of these locations?  Just to clarify, my two
> locations don't interface with each other.  They are two different
> clients of ours each connecting only to the PSTN (one via IAX, the 
> other
> via SIP).
Nope.  No VLANs for us.

>  I should also mention that the "cutting-out" is not an issue
> all the time, it is intermittent and I haven't been able to pin it down
> to one thing in particular (ie it has been reported with everyone in 
> the
> office and on the phones, as well as on a Sunday when only the owner 
> was
> in the office and using his phone.
I'd have to say that our problem is definitely worse when traffic is 
heavier, but I have had drops at off-peak times, too.

- Noah