[asterisk-users] network design philosophy and practice

[Bill Michaelson]

Alex Balashov wrote:
> Send asterisk-users mailing list submissions to
> 	asterisk-users at lists.digium.com
> I'm pretty sure they meant two logical networks.  At least, I hope they did.
>   
Unfortunately, I was indeed referring to two physical networks. Cabling, 
switches, everything, all the way back to the TDM connection to the PSTN.
> David Gibbons wrote:
>
>   
>> Two separate networks? Did I miss something? I feel like I'm taking crazy pills! Two separate physical networks means twice the hassle, twice the maintenance, twice the cost, twice the headache. Not to mention the fact that the whole idea of VOIP is to simplify IT and focus on converging data and voice networks.
>>
>> This is what VLANs and QOS do best. I dare say it's what they were designed foe. I can't think of any reason that I would ever recommend two ports per desk to support telephony -- ever. It's ludicrous to think that two ports will be better than one if we're setting up our VLANs and QOS properly. A phone takes very, very little bandwidth away from the desktop and a decent one will support tagging its frames for the alternate voice VLAN.

I agree, especially about QoS design intent. But I posted my question as 
a sanity check, and there seems to be no shortage of opinions. Now mine:

I can think of two valid reasons to physically segregate the networks:

1) Insurance. I.e., to eliminate the possibility that otherwise properly 
configured QoS mechanisms become broken, either by accident, 
incompetence, or badly-designed or rogue software or hardware - or are 
otherwise handled carelessly as Jerry Jones suggested. But this is not a 
compelling argument to me in any but the most critical scenarios such as 
public-safety applications, etc.

2) Customer preference. If you need the business, then the customer is 
always right. You might not have adequate credibility with the customer 
or influence over the design decision, and if a customer in such a 
situation gets it in their heads that voice and data can't coexist on 
wires, then it can't.

There is a variety of opinions, but no general consensus about where QoS 
failures typically occur, when they occur.

I'm wondering if anyone has anyone has ever experienced QoS issues 
caused by contemporary Polycom phones like IP330s that had workstations 
hanging off their builtin switches? If you did, were you able to 
identify the cause, and was it due to any inherent failure of the phone, 
such as not marking packets or prioritizing dispatch correctly?



-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3234 bytes
Desc: S/MIME Cryptographic Signature
Url : http://lists.digium.com/pipermail/asterisk-users/attachments/20081029/fae2b88c/attachment.bin