Fabric Wars: Too Soon To Pick A Winner

While the networking giants--Cisco Systems, Brocade and Juniper Networks--battle it out for domination of the emerging fabric networking segment, a host of new competitors are nipping at their heels. Other networking companies--such as Avaya, Enterasys and Alcatel-Lucent--as well as computer companies edging into the networking space--such as HP and Dell--are among those joining the fray.

May 1, 2012

6 Min Read
Network Computing logo

While the networking giants--Cisco Systems, Brocade and Juniper Networks--battle it out for domination of the emerging fabric networking segment, a host of new competitors are nipping at their heels. Other networking companies--such as Avaya, Enterasys and Alcatel-Lucent--as well as computer companies edging into the networking space--such as HP and Dell--are among those joining the fray.

One analyst says the rivalry is so strong because each company is offering a largely proprietary approach to fabric and each wants to be the first to dominate the market. Juniper’s QFabric, for instance, is criticized as an expensive vendor lock-in play, still more a marketing plan than an actual product, though Juniper disagrees.

Fabric computing refers to networking technology in which a series of switches are controlled by network intelligence to be operated as one virtual switch, creating multiple paths for data to take through those switches. Fabric enables east-west traffic between and among switches and servers on the same network layer, in addition to the traditional north-south path among the core, aggregation and access layers.

While vendors differ in their technology and approach to market, maximizing east-west traffic "would be the single lighthouse they’re all trying to row to," says Zeus Keravalla, principal analyst with ZK research.

Fabric is essential to handling the massive increases in data workloads created by the explosion of virtualization and cloud computing. Fabric generally is based on one of two underlying technologies: shortest path bridging, a standard approved by the Institute of Electrical and Electronics Engineers (IEEE), and Transparent Interconnect of Lots of Links (TRILL), a standard approved by the Internet Engineering Task Force (IETF).

But because fabric is relatively new, it’s hard to compare different vendor offerings. Dell’Oro Group, a research firm that specializes in the networking space, doesn’t yet track sales or market share numbers for fabric computing. "The industry really hasn't been comfortable defining what really is and isn't a fabric and how to count them," says senior director Alan Weckel.

Juniper has sought to distinguish itself from other vendors with technology it calls QFabric, unveiled in February 2011. QFabric has three basic components: the QFX 3500 line of 10 Gbits per second (Gbps) top-of-rack (ToR) switches, which began shipping in March 2011; an appliance called a Fabric Interconnect that links all the ToR switches; and Fabric Director, a server-based controller for the combined network that gives it that "one virtual switch" capability.

Since that introduction more than a year ago, Juniper has faced criticism from competitors and some analysts that with QFabric, as Keravalla put it, "their marketing was well ahead of product." Cisco also posted a video on You Tube making fun of Juniper’s inability to deliver QFabric, comparing Juniper to a pizza delivery service.

Juniper’s problem with QFabric is that it is a big change, requiring a substantial financial commitment for a customer to adopt it, says Eric Hanselman, networking research director for 451 Group."It’s a more complicated arrangement now," he says, compared to other fabric architectures. While QFabric delivers significantly lower latency than other fabric solutions, it’s a more involved installation than those of competitors. "You’ve got to be willing to make the jump to this non-conventional way of building a switch infrastructure," he adds.

But Juniper explains that QFabric is not an all-or-nothing proposition, nor is its adoption as weak as critics say.

The QFX 3500 switches began shipping in March, but the Interconnect and Fabric Director components didn’t become available until late September of last year, says Dhritiman Dasgupta, senior director of product marketing at Juniper. The switches can operate as standalone switches, but it is only when combined with the Interconnect and Fabric Director that it becomes a QFabric installation--a fabric of up to 6,144 ports. That being the case, QFabric as a whole has been available only for about two full quarters.

Because a customer can add QFX 3500 switches to their data center and then add the other components to make it QFabric, "it provides a very nice migration path," says Dasgupta.

"We don’t come in and say you have to buy this 6,000-port fabric to prepare for this very large environment," adds Denise Shiffman, a VP of product marketing at Juniper. "When you hit a certain amount of [switch] density you can move that into ... the fabric and grow that fabric incrementally."

And despite critics’ claims that QFabric is all pitch and no close, it does have some paying, and named, customers, she says. Juniper lists Sabey Data Centers, a hosting service provider, and Bell Canada, Canada's largest telco, as users of QFabric, plus others that are confidential. Shiffman acknowledges that these are partial deployments of QFabric and not in production environments. The company also has several customers running just the QFX switches for now, "[though] they see it as an on-ramp to the rest of the QFabric family," she says.

Nonetheless, once you go full QFabric, it’s hard to go another way, notes Jason Nolet, VP of data center and enterprise networking at Brocade. "Where you might see Juniper QFabric as being a highly proprietary interconnect ... we try to remain as open standards-based as possible."

Brocade, like industry leader Cisco, bases its fabric technology on TRILL, and began shipping its VDX line of switches in late 2010 and early 2011. Its secret sauce is "virtual cluster switching," which provides the ability to connect a number of VDX switches into whatever arrangement the customer chooses."The customer can effectively flatten their network by combining access and aggregation into this one cluster switch layer," Nolet explains. Fabric flattens the hierarchy of three layers--core, aggregation and access--that is the architecture of client-server networks of old.

Lastly, Cisco aims to maintain its networking lead in the fabric era with its Cisco Unified Fabric "vision," says Shashi Kiran, senior director of Cisco Data Center Solutions. The Unified Fabric strategy combines Cisco Nexus switches and MDS SAN switches with its NX-OS operating system and FabricPath, its variation on the TRILL standard also supported by Brocade. FabricPath has been deployed by more than 1,000 customers worldwide, the company says. The key to Cisco’s Unified Fabric vision is "architectural flexibility," explains Kiran.

"The underlying infrastructure customers invest in has to be adaptable enough to accommodate heterogeneous workloads and diverse business requirements," he says, whether in traditional enterprises, Web 2.0 environments, big data operations, or the cloud or service provider markets.

The shortest path bridging (SBN) protocol is actually the more mature of the prevailing fabric standards, says ZK Research’s Keravalla, and is the technology behind fabric technology from vendors such as Alcatel-Lucent, Avaya and Huawei.

It’s because fabric is just emerging widely that companies are vying against each other for buyer attention, says the 451 Group’s Hanselman. "They are still fundamentally proprietary," he says. "The reason why they are all [going after] each other is because they want their customers to commit to their proprietary fabric."

To separate the hype and marketing buzz from available, reliable and affordable fabric solutions, Keravalla suggests taking a cue from compute vendors on how to choose. The traditional way of looking at network infrastructure has been to look at things like backplane capacity and overall port density. Now, in some ways, that doesn’t matter because the available options are all pretty fast; perhaps server-to-server latency would be a better yardstick.

However Keravalla wonders how best to measure the comparative merits of fabric offerings: "I just think we’re a little too early in it right now to have any kind of good best practice."

Learn more about Research: 2012 Network Computing Salary Survey by subscribing to Network Computing Pro Reports (free, registration required).

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights