[VoiceOps] Enum for carrier services

[Alex Balashov]

Darren,

On 02/21/2011 02:33 AM, Darren Schreiber wrote:

> I am researching whether or not ENUM is a viable solution for
> carrier-grade infrastructure routing. It has been suggested that
> utilizing DNS for number lookups within an internal network is a
> very powerful, redundant way of handling large volumes of traffic
> while also gaining caching and many years of DNS infrastructure
> reliability in one shot.

So, just to be clear, are you talking about the use of ENUM for
large-scale routing that is nevertheless internal to a single
organisation/network/autonomous system/whatever you want to call it,
or for inter-domain VoIP peering?

The reasons for why the latter hasn't so much taken off are mostly
rooted in politics and economic pragmatism.

> Things is, I can't figure out why a 10+ year technology doesn't
> have excellent penetration if it's this good. I can guess that
> hooking a bunch of DNS servers to potentially SQL-based DNS
> servers has the argument of "why not just go direct to the DB to
> reduce latency/complexity" but DNS provides other advantages,
> ranging from local caching on each box (comes "free") to inherent
> redundancy with multiple name servers (though there's a delay in
> failing over obviously). And of course, if done properly with
> partitioning, DNS can be done via flat-files for ultimate speed.

Everyone has a different opinion on this, but I would say two things 
matter the most:

1. The built-in freebies (replication, delegation, redundancy, etc.) 
can also be achieved with databases fairly easily and more 
straightforwardly, while achieving economies of scale and leveraging 
existing database infrastructure since non-trivial networks depend on 
highly available, redundant database clusters anyway.

This is often seen as a superior alternative to setting up an entirely 
parallel DNS-based infrastructure just for routing.

But even more importantly:

2. VoIP interconnection still being the largely PSTN-centric world 
that it is today, in the sense that by far the primary preoccupation 
of any non-Tier 1 service provider is still exchanging calls with the 
PSTN, not with other ISPs, the requirements of the routing logic are 
more complex.

There's almost certainly heavy business-layer  cost routing (LCR), 
inter vs. intra-jurisdictional routing and other billing 
considerations playing a key role.

That doesn't play well with the relative simplicity of what DNS is 
designed to return.  Sure, in the end it comes down to "route the call 
to this IP address," but the method by which I must arrive at that 
decision entails consulting tariffs, rate decks, etc., which means 
dipping a database anyway.  This is true even for deciding which edge 
element to run traffic out of in a purely internal scenario, at least 
for any reasonably complex infrastructure.

With enough spiritual commitment and devotion, I suppose one could 
plumb all that through DNS as an end-stage interface, but that's just 
an extra--and fairly complex--layer of intermediation.  It doesn't 
really do anything but add extra network elements and systems purely 
for the sake of doing so.

ENUM would make a lot more sense in some sort of post-PSTN environment 
where voice is just another application framed over dumb pipe, much 
like instant messaging or HTTP.  In that paradigm, it really is just 
about figuring out which IP endpoint to send the message to, which is 
what DNS is good for.

That may happen someday, but today the process is dictated by all 
sorts of regulatory, financial, security and policy considerations 
that require a muscular backend whose fronting by DNS would be kind of 
contrived and forced.

Or so goes my account, anyway.

Cheers,

-- Alex

-- 
Alex Balashov - Principal
Evariste Systems LLC
260 Peachtree Street NW
Suite 2200
Atlanta, GA 30303
Tel: +1-678-954-0670
Fax: +1-404-961-1892
Web: http://www.evaristesys.com/