VOIP provider
Bill Woodcock
woody at pch.net
Wed Aug 3 09:08:30 UTC 2005
On Tue, 2 Aug 2005, Shane Owens wrote:
> 1: Does it make sense to scatter nodes around the globe to limit latency on intraregional calls? If so how many? We were
> thinking about 7 placed at strategic points around the globe.
The short answer is "yes". This is a VoIP peering issue, which is
basically just like IP peering, but higher up the stack. There will
actually be a VoIP Peering BoF here at the IETF later this afternoon, and
it's been the subject of a lot of discussion.
To give you a concrete example of why local gateways are needed, I have
offices in San Francisco, and we tried a VoIP gateway provider, once,
which located its _single_ gateway in Florida. So all of our "local"
calls to PSTN numbers in California went to Florida across the Internet,
before returning to California. The latency isn't that bad by itself, but
combined with the carrier's mediocre bandwidth, it made for very serious
voice quality problems. We wound up putting up our own PSTN gateway in
San Francisco, and we divide calls between that (California calls) and two
different VoIP carriers (everything outside California, based on price).
If the VoIP carrier had had gateways on both the east coast and the west
coast, they'd have all of our business right now, because we could hand
traffic off to them at PAIX or 1 Wilshire or the SIX, and all would be
good. But they ignored the underlying infrastructure, to their detriment.
> 2: Is a softswitch architecture preferred to a proxy server/Media
> Gateway (Vonage) only type architecture?
You need both.
> 3: What protocols should be used for firmware upgrades to ATA
> devices? We are thinking HTTPS or SFTP, or HTTP if those aren't
> available on selected devices. I am trying to stay away from TFTP
> for security reasons.
What security risk does TFTP pose that isn't also shared by HTTP?
-Bill
More information about the NANOG
mailing list