The Best of Both Worlds IP switching promises to combine the advantages of switches and routers

So carriers and Internet service providers are embracing Internet protocol switching with the same fervor Americans have exhibited in adopting the Macarena dance. Once again, however, as a new technology is introduced to the market, vendors can't seem to find a common ground, making the technology choices as difficult to decipher as the lyrics to the Macarena song.

Ipsilon coined the term IP switching earlier this year when it announced it would integrate routing and switching onto a single technology platform to run over asynchronous transfer mode. Cisco Systems and Ascend Communications were quick to follow, each announcing its own strategy at the Networld+Interop show in Atlanta last month (Table 1).

The interest in IP switching stems from its promise of delivering the speed of switching with the intelligence of routing. ISPs and carriers are dealing with unprecedented traffic levels on their networks because of the increased number of Internet users. The amount of traffic being routed and switched between sites is not being effectively managed, and bottlenecks are occurring, especially in ISP networks. Analysts expect that carriers will begin experiencing networking troubles in six to eight months.

Some simple trends are emerging, such as more wide area network switch deployment in ISP networks to strengthen their core or more router deployment in carrier networks to support growing Internet services. By most industry watchers' standards, though, adding more switches and routers will not be enough to keep up with the increased demand put on carriers. A new technology, most likely IP switching, is needed to guarantee fast, reliable service with low latency. IP switching is expected to be the answer to more integrated, easier-to-manage networks.

A common goal Networking vendors are all trying to solve the same problem, says Michael Howard, president and co-founder of Infonetics, a San Jose-based consultancy, adding that today's routers are too slow for large-volume networks. The latency is causing traffic jams along the Internet that will spread into the public infrastructure, he says.

"IP switching will play an increasingly important role in carriers' networks in the next few years," says Fred McClimans, president of CurrentAnalysis, an Ashburn, Va.-based consultancy.

"We are seeing a shift from ATM-only networks to a hybrid integrated IP and ATM solution," McClimans says. This shift is occurring because ATM standards are not fully defined and ISPs are more inclined to extend their already IP based networks, he says. "Why not give [ISPs] something that will not require a major overhaul of their network?" McClimans asks.

Since Ipsilon's introduction of its IP switching concept in March, five WAN switch manufacturers have signed on to test and implement the technology and a total of 20 vendors participated in a switch-off in August (see sidebar on page 28).

Ipsilon's protocols have been recognized by the Internet Engineering Task Force. The public domain specifications from IETF include the Ipsilon flow management protocol, described in request for comment (RFC) 1953, and the general switch management protocol, RFC 1987.

Ipsilon's technology identifies flows of traffic cutting through IP switching over an ATM network. The Ipsilon approach uses not ATM signaling but IP signaling to speed up the switching process. This is what differentiates it from approaches used by Cisco, Ascend and Newbridge Networks.

Ipsilon's IP switching promises to offer ISPs and carriers cost benefits but still needs quality-of-service (QOS) parameters to be defined, says Larry Blair, vice president of marketing at Ipsilon. QOS is a measure used by most state public utility commissions to determine if carriers are providing their customers with satisfactory service. Traditionally QOS was designed for voice service, but it is now used by carriers for most of their service offerings.

At Networld+Interop last month, Sprint and Fore Systems announced plans to develop QOS parameters for Ipsilon's GSMP RFC 1987 that would make the protocol suitable for carrier deployment (Telephony, Sept. 23, page 7).

Sprint became the first carrier to publicly reveal that it would invest time and effort to ensure an IP switching technology would meet the needs of carriers. Sprint sees the benefit of IP switching as a cost saving and technological solution for traffic congestion and bottlenecks, especially for its IP network services, says Dom DeAngelo, vice president of product management at Sprint.

The company is in the early stages of looking at IP switching and is not yet committed to any particular vendor, he says. Deployment plans are six to eight months down the road, adds DeAngelo.

For the first time, the efforts of Sprint and Fore will allow carriers to guarantee IP services from the central office to the customer site, says Tim Wilson, senior consultant at Decisys, a Sterling, Va.-based consultancy. Other approaches

Cisco Systems recently announced its own variation on IP switching-tag distribution protocol (Telephony, Sept. 16, page 14). The company claims its approach will use the QOS parameters inherent in ATM and its existing StrataCom switching platforms. Cisco's tag switching integrates IP routing and switching onto a card that can be deployed on a Cisco 7500 router or a StrataCom BPX/Axis switch, says Andrew Greenfield, director of service provider marketing at Cisco's WAN switching division, formerly StrataCom.

Tag switching will allow routers or switches to apply tags to traffic flow based on layer 3 routing protocols. The tags are then interpreted and switched based on sub-net information, eliminating the detailed lookups that typically slow down the routing process (Figure 1).

Although Cisco customers will have to wait until mid-1997 for tag distribution protocol products, some carriers may be willing to wait if they've already made a substantial investment in Cisco 7500 routers or StrataCom BPX switches.

Cisco has been accused of trying to hold on to its monopolistic corner of the market with its proprietary approach, but Ipsilon's approach was also proprietary when it was announced earlier this year. Cisco will present its technology to the IETF for formal approval and public distribution, Greenfield says.

Because Ascend Communications tackles IP switching more from a fast-routing approach it will not have to worry about standard approval process like its competitors. The Alameda, Calif.-based company is targeting ISPs exclusively with its recently announced GRF 400 IP switch (Telephony, Sept. 9, page 8). The GRF will allow ISPs to support layer 1 IP switching and layer 3 routing, according to Bernie Schneider, Ascend's vice president of strategic business development. The switching functionality is an off-the-shelf chip set, but the route tables were designed by Ascend (Figure 2).

The GRF has been dubbed a "fast router" by some of its competitors because of the lack of layer 2 switching support that most carriers require. Although the GRF's switching capabilities are not up to par with Ipsilon or Cisco, Ascend has added distributed routing functionality that is expected to let ISP networks grow without the same cramping they are experiencing today.

The routing features were designed by NetStar, a high-speed routing company that Ascend recently acquired. The GRF supports four media cards that each have their own routing table. The GRF is also designed to send IP traffic over multiple WAN connections such as frame relay, Sonet and ATM.

Newbridge Networks is taking yet another road with its IP switching plans. Although the Herndon, Va.-based switch manufacturer has yet to announce specific product details, it has stated it will migrate the ATM Forum's pending multiprotocol over ATM (MPOA) specification to its MainStreet ATMnet line of WAN switches, says John Shaw, assistant vice president of product marketing at Newbridge.

MPOA uses ATM switched virtual circuits, says Eric Andrews, director of product management of Vivid at Newbridge. The technology in effect operates like a WAN router that does wire-speed forwarding, he adds. Newbridge believes MPOA is more suitable to carrier needs because it also let carriers support virtual local area networks services, Andrews says.

Before Newbridge can add MPOA to its MainStreet WAN switches it must first implement the ATM Forum's full suite of switched virtual circuit specifications and the forum's private network-to-network interface, Shaw says. This is expected to be completed in the first half of 1997.

In the forum, Newbridge is one of the proponents of the MPOA specification and to date is the only vendor that has products based on the interim specification. MPOA was originally designed with the LAN in mind, but Newbridge believes its place is also in the WAN.

While Ipsilon, Cisco, Ascend and Newbridge are only a handful of vendors with IP switching strategies, many others are expected to support the technology soon.

General DataComm will roll out support of Ipsilon's IP switching technology on its Apex line of ATM switches by the end of the year. And Bay Networks also alluded to IP switching products but is more reluctant to say which approach it will adopt. Although Cascade Communications has not made any formal announcements about its IP switching strategy, officials from the company did say that Ipsilon's and Cisco's technologies wouldn't fare well in carrier networks.

"Right now, neither tag switching nor Ipsilon [switching] solves QOS issues," says Bob Machlin, vice president of marketing at Cascade. "Fundamentally, Cisco and Ipsilon both have good plans to make IP networks go faster, but neither offer a connection-based backbone needed for end-to-end QOS."

Before the end of the year, Cascade will roll out IP switching plans that will be designed with carriers' networking needs in mind, Machlin promises.

Analysts agree it's too soon to determine which IP switching technology will prevail with carriers, but it is clear that IP switching will play an important role in carriers' network architecture strategies.

"It's like trying to compare a motorcycle with an automobile," McClimans says. "Both get you to from point A to point B, but in different ways." Carriers will make decisions on which technology to deploy based on interoperability, the different types of traffic they need to support and what their installed base looks like now, he believes.