IPNet forerunner to new network

IPNet has become something of a dirty word in the current tussle over Internet access infrastructures, but for Telecom's Graeme Rowe, the dedicated IP network is version 1.0 of technology that will change the way Telecom - and other telcos worldwide - will do business in the future.

At the heart of the row over Telecom's plans to charge home Internet users 2 cents a minute if they don't use an 0867 access number is the telco's IPNet service - to which its own ISP, Xtra, can offer dial-up charges cheaper than competing ISPs.

IPNet is, in essence, version 1.0 of technology that will change the way Telecom - and other telcos worldwide - will do business in the future.

Telecom manager of development data Graeme Rowe describes IPNet as a service placed in one corner and connected to the PSTN (public service telephone network). It is a matter of time before IPNet and some of the PSTN are merged into what is developing as one very large, intelligent switch.

It is a time of change for IP (Internet protocol), which is being asked to do things it was not designed for, he says.

Historically, telco services have been split into what's called deterministic and futuristic. A few years ago telcos became entrenched in ATM [asynchronous transfer mode] on the deterministic side - meaning that guaranteed levels of service could be offered.

"But ATM became too expensive and the IP world has begun moving into the ATM space," Rowe says.

"By the end of 1998, IP was obviously going to win because of cost." (That's exemplified by the sudden rash of mergers among large ATM vendors.)

"IP has had to change to be more deterministic," he says. "ATM had promised to send information with a very high guarantee defined by SLAs [service level agreements] and priorities of traffic."

In 1997, the Internet 2 Group faced two problems: how to make IP more deterministic and the fact that there were not enough IP numbers.

Version 6.0 of IP - currently version 4.0 is used - was going to define quality of service and expand numbering so there were enough for everyone on the planet.

But out of the quality of service work which was done, it was discovered that it was possible to tag frames. In TCP/IP today there are now a series of bits that can be user-defined by a vendor.

"Some of the work on version 6.0 was transported to TCP/IP to allow prioritisation - allowing 16 levels of priority," Rowe says.

Is Telecom doing it? Yes and no, he says. "At the moment it's proprietary to each vendor. The next step is to find out how you allocate priorities."

That introduces another element. Among others, but probably leading the market, is a partnership between Cisco and Microsoft to develop a directory enabled networking (DEN) standard.

"The directory structure will be used to define the user with a profile generally based on security access of applications, then linking the applications to priority use within the network. It's probably a year out from a useable version. Then we have to rebuild the network to recognise it."

There are currently two parts to the network: the public part, based on the Internet; and private networks, based on frame relay or ATM. They are individually connected to the user via routers.

"Increasingly, the private network is starting to look like a big router," Rowe says.

"You can now take some of the complexity and put it in the network; in theory, it is shared by many users so the incremental cost to the user goes down.

"Our network plans are to look at the architecture and the type of intelligence needed for prioritisation. Underpinning the IP part of the network is ATM, which will be around for some time. The intelligent IP lives at the service layer.

"As you begin to provide layer three capabilities in the private network, you use layer two protocols. It's getting very blurred between layers two and three. The old OSI [open systems interconnection] model goes out of the window."

The fastest growing data service product for Telecom is Lanlink, which was built by the former Netway. TCP/IP packets fit within the frame relay.

"The next generation of network evolution will be adding more layer three IP capability," Rowe says.

Benefits to the user comes down to a cost equation. Because a lot of functionality will reside in the network, the cost of a router will fall from around $7000 to $1000, he says.

"With new intelligence and flexibility in the network, how you get to the customer becomes irrelevant. It's just a cost equation within a range of connectivity options."

There are plans to integrate the core of the PSTN into the new architecture, he says.

That's for functionality gain, such as linking to the white pages through an icon.

Timing is down to vendors such as Cisco and Microsoft.

"As soon as the vendors build the technologies, we will test and trial them."

It may not be that far away. In the US, AT&T already has an SLA-based IP service on its Web site.

"AT&T is a variation of what we're doing with IPNet and Lanlink," Rowe says.

Clear refused to talk about its plans on the grounds of commercial sensitivity, while Telstra didn't respond to phone calls.

Join the newsletter!

Error: Please check your email address.
Show Comments
[]