• United States

What the heck’s an inFRAnet?

Sep 27, 200412 mins
MPLSNetworkingWeb Development

Service providers view these next-generation business-class IP networks as a means to re-create themselves. This time, are corporations ready?

If carriers were to create a business-class public IP infrastructure with guaranteed QoS, reliability and security, would corporations willingly pay to run their most demanding applications on top of it?

A recently formed industry group is betting on it. More than two-dozen vendors and service providers have banded together to create such a public IP infrastructure, dubbed infranet. The Infranet Initiative group includes global service providers such as AOL, British Telecom, Deutsche Telekom, France Telecom, Level 3 Communications and Qwest; network gear vendors Ericsson, Juniper, Lucent and Polycom; and application and computing companies HP, IBM and Oracle. Enterprise organizations are noticeably absent from the roster.

The Infranet Initiative is Juniper’s brainchild – a way for the company to help its customers, “rudderless” carriers, find direction, says Pradeep Sindhu, Juniper CTO. Juniper realized it would need to get carriers on a new course, one that took them away from their long-practiced “one application/one network” way of doing business “that’s too costly and complex to persist for the long term,” he says.

The ubiquitous, cost-effective Internet, on the other hand, provides the right, scalable foundation, Sindhu says. But for carriers looking to expand their premium service offerings, the Internet is too limiting in the types of applications it can support. It lacks the stringent QoS, reliability and security mechanisms needed for premium business services, he adds.

Applications that could benefit from the infranet approach include enterprise-wide VoIP requiring handoffs from one carrier’s network to another’s, inter-company peer-to-peer collaboration, utility computing and multi-provider VPNs, say members of the Infranet Initiative’s leadership council (IIC).

A multi-provider VPN service would be ideal for a business traveler whose office is his PC, says Marco Limena, vice president of HP’s Network and Service Provider Solutions unit and IIC member. No matter whether the traveler lands in Asia, Europe or Latin America, he would use the same procedures to access an IP network and would receive the same connectivity speeds, connection quality and security worldwide, he says.

With an infranet, a company would be able to contract for that consistent multi-carrier VPN service. Using specifications under development by Infranet Initiative members, carriers would be able to hand off calls seamlessly among themselves while a corporation would receive one bill and the user gets the same service characteristics – availability, cost, QoS – from anywhere he lands, Limena says.

Industry déjà vu

If this goal sounds familiar, it is. In the late 1990s, ISPs drummed up the concept of business-class Internet upon which they could offer premium services.

That effort failed for a few reasons, the IIC says. One was timing. “Some businesses still viewed the Internet as a fad that would fade with the bursting bubble, plus the Internet wasn’t yet the source of viral disease and hacking that it is today. So the problem being solved wasn’t as acute as it is now nor were the applications that could use it as prevalent or important,” a council member says.

What’s more, underlying standards – mainly IPv6 – were not sufficient enough to make the promise of a business-class Internet a reality. That’s still a problem today – hence the Infranet Initiative. “The driving need is largely the same and growing each year,” the IIC says.



  • Do you support the Infranet Initiative?
  • What is the status of MPLS technology within your backbone?
  • What MPLS services do you offer?
  • Does the provider of your MPLS gear belong to the Infranet Initiative group?
  • Do you interconnect your MPLS backbone with another carrier’s MPLS network?
  • Can I run multiple applications at different priorities — with different QoS parameters associated with each priority ­ across the same MPLS connection?

In the early 1990s, telecom carriers also tried to parlay their packet-based infrastructures into business-class, revenue-generating platforms for virtual services. Back then, carriers viewed frame relay and ATM as the be-all and end-all infrastructure technologies. They heralded ATM for its advanced QoS features.

But frame relay and ATM were stymied by their connection-oriented natures, says Christine Heckart, vice president of marketing with Juniper. Infranets have the big advantage of relying on routed Multi-protocol Label Switching (MPLS) backbones, she says. With MPLS, IP traffic can be steered over a variety of routes to enable a particular class of service or guaranteed service level. By now, most mainstay carriers have committed to MPLS on their backbones and are ready to capitalize on their new mesh architectures and subsequent routing flexibility.

The MPLS backbone

Heckart points to start-up carrier and working group member Masergy Communications as already providing infranet-type services on an MPLS backbone. Masergy’s inControl IP services let a corporation run multiple applications at different priorities – with different QoS parameters associated with each priority – across the same connection.

If Masergy’s success is any indication, the Infranet Initiative is definitely on to something. Since launching in 2000, Masergy has amassed more than 200 enterprise customers in 30 countries, the company says. Among them are Fluor, Genesis Microchip, GMAC Commercial Mortgage, Hallmark Channel, LifeCare Hospitals, Tampa Armature Works and The Weather Channel.

At Tampa Armature, voice over the infranet-style MPLS backbone is proving far more reliable than voice over frame relay, says John Sarmanian, MIS manager at the Florida electrical repair and manufacturing company. The company had converged voice onto its frame relay network about three years ago to save on long-distance charges, but suffered a nightmare of infrastructure problems as a result. Despite carrier assurances to the contrary, the 256K bit/sec circuits didn’t provide enough bandwidth for voice and data, and the company’s routers weren’t capable of supporting necessary QoS mechanisms, he explains.

Who’s who in the Infranet Initiative group

More than two dozen service providers and vendors are working to create an alternative public IP infrastructure.Click here for more.

“So we took the site I had the most trouble with and put in a Masergy circuit for a two-month trial. All the problems we had been experiencing with voice over frame relay went away,” Sarmanian says.

Tampa Armature now runs voice and data between its headquarters and 10 remote locations. It has a centralized T-1 for Internet access and 3M bit/sec pipes for the converged traffic, but pays less monthly than it did for the 256K bit/sec circuits. “We’ve been very satisfied with the QoS for voice. We’ve been getting 100% packet delivery with less than 10 millisec of jitter – and the Masergy network has never been the source of a problem,” Sarmanian says.

This infranet-style service has been so flawless that Sarmanian has begun adding video into the mix to reduce travel for in-person meetings. And he’s renewed his service contract for two years – even though the previous paperwork had not expired.

An infranet-style MPLS service has proven valuable for Vienna, Austria, oil company OMV, says Ulf Busch, CIO. Busch contracts with T-Systems International, the Deutsche Telekom unit serving enterprise customers in the U.S. and elsewhere, for the carrier’s IntraSelect MPLS IP VPN service.

Application performance has improved across the board over the company’s previous Layer 2/3 network, Busch says. Response times are lower for the company’s SAP application, voice quality has increased, intranet access speed is faster, and Internet access has been centralized, he says. Busch says he’d evaluate full-fledged infranet services as the need arises. Such a service would use IIC-developed and IIC-sanctioned open standards for accessing the infranet service and carrying traffic from one network to another while delivering the same level of quality and security to corporations, all on one bill.

In theory, carriers such as Masergy and T-Systems could link their MPLS networks today to provide infranet-type services across their networks. This was the goal of the now-defunct CoreExpress and a few other start-ups that launched early this decade with the promise of offering service-level agreements that extended across multiple ISP networks. But the reality is that achieving such partnership has proved prohibitively costly and cumbersome, says Jody Craft, executive vice president of T-Systems. If the Infranet Initiative accomplishes its goals, that problem would disappear. “With all of the triage stuff – billing, security, etc. – pre-defined and supported by vendors, we could reduce the time it takes for an interconnection from two years to two months,” Craft says.

Web services for intelligence

Just as MPLS will make the infranet, so too will Web services. “The selection of Web services as an element in a new IP architecture is critical because Web services is explicitly a tool for integrating computer intelligence into networks,” says Thomas Nolle, president of consultancy CIMI, in an infranet white paper called “Infranets: Fulfilling IP’s Promise.”

The IIC hopes infranets will provide the global infrastructure necessary to support on-demand computing and applications that take advantage of grids (and hence the reason HP, IBM and Oracle participate), Juniper’s Heckart says. The IIC intends to rely on Web services standards for signaling and security, for example. It will use Simple Object Access Protocol, the XML-based scheme for allowing networked computers to talk to each other as part of a Web service; and WS-Security for securing SOAP communications, says Kevin Dillon, director of strategic development at Juniper.

“The edge of an infranet is a ‘trust barrier’ where users are admitted based on their validated identity and right to claim service access,” Nolle says in the white paper. This separates the management of a network as a community from the management of a network as a connection/transport resource, and this division of functionality has major impacts on the nature of services in an infranet.”

So for that multi-provider VPN scenario, instead of creating a VPN for each customer or service type, an infranet creates a network partition that would let carriers divide traffic according to QoS and security requirements. Carriers can put multiple customers and services on one partition if the requirements are the same and trust has been established, Nolle explains.

For application and content services, a server would be admitted to the infranet as network member. Or, it would be admitted as a provider of Web services whose applications are published in service registry and made available to network users.

“This means that infranets can accommodate both the legacy model of distributed computing and the new service-oriented architecture of Web services,” Nolle says in the white paper. He suggests that, because the infranet architecture will provide service interworking, legacy frame relay and ATM services could be joined in an infranet with an IP services offering Web services applications.

The use cases

The IIC spells out the service interworking and other technical aspects in a three-level reference architecture. The IIC intends to incorporate standards work already accomplished within the IETF and International Telcommunication Union where possible, council members say.

The IIC plans on proving the validity of this architecture by mapping a half-dozen or so use cases against it. In progress are multi-provider corporate VPN service, including performance assurances; software distribution and maintainence; Web radio; and fixed/mobile convergence. Other use cases the IIC have discussed include multi-provider Session Initiation Protocol voice calls; per-session billing arrangements; peer-to-peer collaboration between corporations; and multimedia home gateway/home network. All these applications are among those that the IIC says cannot be handled adequately by today’s public IP services.

The IIC would like to move the use cases into pilot tests within six to 12 months, Dillon says. Its endeavors should prove worthwhile, industry analysts say.

“This is an important development for the industry because prior to the Infranet Initiative there really was no managed effort to bring about some of these changes in the public Internet,” says Mark Bieberich, an analyst with The Yankee Group.

Even for carriers with long histories of turf wars and proprietary mindsets, the collaborative approach of the Infranet Initiative is a welcome change. “The standardization efforts for the infranet right now are all about how technically we can do such services. Nobody’s ever taken this service-creation approach before,” T-Systems’ Craft says.

And as HP’s Limena says, “Shame on us if we aren’t successful. This is an opportunity to differentiate from the public Internet’s best-effort approach with a new architecture that offers the highest-level, premium user experience.”

Packet handling stratum: A sub function to each of the stratums above it, the packet handling layer handles incoming and outgoing packets – appending labels to them, encapsulating them in MPLS headers, queuing them, etc. Some “processing” of packets, such as encryption, intrusion detection and anomaly monitoring, also takes place here.

Network policy and control stratum: At the highest level, resource allocation tracking takes place here to ensure that the network can meet the reliability, security and availability commitments related to a requested service. Policy exchanges also occur here, to check the request for resource consumption against business policy guidelines. The group is still working on the specifics of the interface between the packet handling and network policy and control stratums.

Service signaling stratum: Here is where, through protocol exchanges, the needs of a particular service are communicated to the network policy layer. The communication can be initiated from a network client, with a client being any origination or termination point such as a user’s PC or a network edge router. The signaling layer will use Web Services Description Layer and other Web services standards.

*Policy-to-Element Interface   ** Signaling-to-Network Interface