MSOs Seek Closer Ties to 'Peers'


In the Internet world, your peers are your partners — and they can save a good amount of money and data hassle.

So it isn't surprising that as MSO high-speed data services gain volume, operators are looking to forge peering agreements with other network operators, to cut transport costs and improve network performance.

While these agreements are modestly beneficial for data traffic at present, operators foresee a future in which peering deals could power voice traffic that never touches the public switched-telephone network run by their competitors.

Essentially, peering agreements allow two operators to create a common interconnection point and flow traffic from one's network to the other's.

While smaller operators pay for transit connections to links to major backbone carriers such as WorldCom Inc. or Sprint Corp. for long-haul data transport, regional network operators with roughly the same traffic can sign a "wash agreement" in which no money changes hands.

It's not much of a stretch to see that with these agreements, operators can cut down on the number of paid transit connections they need to buy from major backbone carriers.

Also, by bypassing the connection-tangled Internet, they can improve the speed at which data packets are sent.

But the process is slow and negotiations, driven by the nuances of network size, can be lengthy. So former Excite@Home Corp. affiliates building their networks from scratch have been buying connections to get started.

Only last month did Cox Communications Inc. activate its first peer link with an undisclosed partner, according to vice president of data engineering Jay Rolls.

"For every [megabit per second] of data we exchange with a peer, it is saving us between $100 and $300 [monthly] per meg. It is substantial," Rolls said. "We are doing something like six or seven [gigabits per second] per month peak from the Internet."

Comcast Corp. has not quite started forging peering agreements, but "obviously that is something that we would take a look at — that's just pretty straightforward a cost issue," said Mark Coblitz, senior vice president of strategic planning. "Are you sending the traffic to a third party, or can you keep the traffic first of all yourself, and secondly, within the cable networks?"

For former Excite@Home affiliates, this is a relatively new concept, according to Michael Harris, president of cable technology analyst firm Kinetic Strategies Inc.

"To a certain extent, that was what @Home was supposed to do and did," he said. "Now that all of the operators have been forced to do their own thing, to the extent that they could find a way to share traffic among themselves and not have to move to external third parties, there is a cost savings there."


The other key benefit to peering agreements comes from the performance improvements in avoiding the sometimes-wacky routing typical on large backbone networks.

"This is part of the nature of the game that sometimes the shortest route between New York and Boston is by way of Santa Clara," said Jupiter Research Inc. senior analyst Matthew Berk. "Typically, when you go through the larger providers that have their own backbones — especially if you are a relatively smaller provider — you don't have any say as to how they route you. So it is hot-potato routing most of the time."

Rolls said the average user won't really notice that their packet of data has taken fewer hops, but such power users as real-time gamers will.

"Although some customers look at the hop numbers and get overly spastic about that, it's not really that relevant," Rolls noted. "What's more important is the latency it represents. And I think it can result in, just shooting from the hip, a 10 to 30 percent latency reduction."

The creation of peered networks also has an effect on the overall Internet grid, shifting some of the emphasis from centralized backbones and distributing it to other networks, Berk said.

"The way you think about it is that we are moving towards a relatively decentralized model, where you can have local peering at points of presence or even not necessarily where there are common points of presence. You get more control, and in terms of network hops, it can be quite advantageous."


With performance and cost benefits, MSOs should be falling over themselves to set up peering agreements, right?

No. As it turns out, the process isn't easy, especially for operators dealing with a larger network with greater traffic levels.

"If I go to a network and say 'I want to peer with you,' they think, 'You know what? We probably are such a big player these guys probably would buy from us for a peer.' That's where all of the funny nuances come in, with the bargaining," Rolls said. "If I go to those guys and I don't meet some of their criteria and I am not well-prepared, they are just going to say, 'No,' right away, and it would be a futile exercise."

With bigger carriers out of range, it is a natural that cable operators of relatively similar size would look to each other.
Those discussions have already begun, and the subject was on the agenda at last week's cable backbone roundtable, which convened at Cable Television Laboratories Inc.

Even among MSOs, the negotiations will be complicated by the question of equivalent traffic.

"It is very straightforward when you are in the telephony world and both ends are IP. And so I think those are different circumstances," Coblitz noted. "There is very likely to be a balance — a relative balance of traffic with telephony. It is unclear whether or not there is going to be a balance of traffic for the pure data services. And that is the basis of a negotiation."

It makes sense that these deals will likely be forged between partners with relatively equal traffic loads, analyst Harris said.

"AT&T [Broadband] is going to generate a ton more traffic than Insight [Communications Co.]. However, they still might be able to negotiate economics among themselves, given the two-way flows they might have between their networks," he said.

"Getting into what you might think of as arbitrage, or figuring out economically how it might work or technically making some changes to the routing architectures and how the networks engineer for each of these guys — that's not trivial."

Even with parity, the challenge is finding a location at which to make the physical link.

One of the reasons why cable operators have been slow to reach peering agreements has been the lack of available spots where networks can be interconnected.

That situation is made easier with the rise of several regional peering points in major metropolitan clusters, which give operators more access to central locations for linkups.

"There is a lot of support and groundswell toward really putting energy to these places, because if you have 20 network providers who are in a certain city, and they all start doing links to each other, the mesh of links would be ridiculous," Rolls said. "So the idea is, 'Why don't we all land one circuit to the same place and just interexchange our traffic there?' "


Telephony is another reason cablers aren't hurrying into peer agreements. While data-peering agreements can save operators some money now, they may prove more valuable in the future, when voice-over-Internet protocol businesses become becomes better established.

With VoIP, MSOs could keep some traffic completely within the cable network, reducing costly voice interconnections to the switched telephone networks run by the Baby Bells — companies that are their direct competitors in the voice business,

But most estimates say there won't be large-scale IP voice rollouts well into 2003, so there isn't exactly a burning need to rush into peering agreements.

"What we want to do is get those peering relationships, and the next step is obvious — maybe we want to be doing some voice-over-IP stuff together, where my subscribers can call your subscribers over IP. That would be great," Rolls said. "I think it wouldn't be that difficult, once voice over IP was kicked off in a serious way at two different operators, for them to exchange traffic it could probably be done within a year."

Coblitz, meanwhile, sees more peering benefit in IP telephony than data, as cablers work through the issues of establishing IP voice service.

In addition to firming the PacketCable 1.2 specifications governing consistent network voice signaling and delivery, "the next issue is going to be how to actually effectuate that, and use it to reduce cost — and ultimately enable new services that require both ends to have similar capability," Coblitz said.

"So that's what the thinking is, though since we are in the very early days the delivery of any of these services, we haven't yet begun the discussions about what those peering arrangements might look like," Coblitz added.

All in all, cable's venture into peering agreements is part of their evolution into Internet and communications services providers.

"I think it is a nice idea," Harris said. "I would be surprised to see them move as an industry at lighting speed on the issue. But I think as voice gets closer, then the idea of an MSO IP packet being able to go from one subscriber to another without traversing a competitive network is attractive."