Will Price Drops Drive 10-GbE to Adoption?
When it comes to networking, speed is usually better. However, the industry has been ahead of customers the last few years during a rollout of 10-Gigabit Ethernet products. Simply put, there just wasn't the need for 10-Gigabit Ethernet implementations on a wider scale, and when you factor in high prices — the kind always paid by early adopters — 10-Gigabit Ethernet was cornered into the server-to-server market.
But this is changing, and many industry observers are saying that 2004 could be the year that 10-Gigabit Ethernet (also referred to as 10-GigE or IEEE 802.3ae) breaks out of the server room and becomes more widely deployed in the enterprise. As 10-Gigabit Ethernet products steadily drop in price, more enterprise are seduced by the notion of installing the technology as a hedge against a future when bandwidth needs will (inevitably) increase. Future enhancements to 10-Gigabit Ethernet will allow it to run over copper lines, making it attractive to implement in current networking schemes, but today's IT department is most likely to install 10-Gigabit Ethernet because of lowering prices.
How Low Will They Go?
How fast are prices falling? Seamus Crehan, a research director at Dell'Oro Group, said in a report that average selling prices for 10 Gigabit Ethernet fell by almost 60 percent in the fourth quarter of 2003, leading to a 288-percent increase in port shipments. According to his research per-port prices have fallen by more than $30,000 since the fourth quarter of 2002.
"Although still a relatively small market in comparison to the more mature technologies such as Gigabit and Fast Ethernet, these dramatic price drops should help facilitate continued strong growth for 10 Gigabit Ethernet," he wrote in a summary to the report. "A 10-Gigabit Ethernet port now offers a discount on a dollar-per-gigabit-of-bandwidth basis over a similar, high-end Gigabit Ethernet port, making it attractive for enterprise applications such as inter-campus, backbone and data center connections."
Leading the way in 10-Gigabit Ethernet shipments: Cisco, Foundry and Extreme. Overall, according to research from The Yankee Group, 10-Gigabit Ethernet shipments grew to $90 million in 2003 after totaling $42 million in gear in 2002.
However, it may be a year or two or even three before 10-Gigabit Ethernet becomes more widely deployed. The market is definitely lagging behind the technology, and many enterprises are just now adopting Gigabit Ethernet, never mind 10-Gigabit Ethernet. A lack of perceived need has prevented a more widespread adoption of Gigabit Ethernet, according to a report issued by market-research firm In-Stat/MDR. Sam Lucero from the Scottsdale, Ariz.-based firm wrote in a summary of his study that "the most significant barrier to adoption of GigE and 10 GigE appears to be the common perception among potential customers of the simple lack of need for a full Gigabit of bandwidth to the desktop, or a full 10 Gigabits of bandwidth anywhere in the organization."
The numbers backed him up: 64 percent of the survey respondents said that the biggest obstacle for adoption was the lack of a perceived need of the high-end networking technology. After that, monetary reasons came into play: over half the respondents said that Gigabit Ethernet just wasn't worth the money and that they lacked the budgets to implement it.
Eyeing the Future
All in all, only 46 percent of respondents said their organizations had implemented Gigabit Ethernet (almost all on the enterprise level), and they did so as part of a plan for to ensure adequate bandwidth in the future, not to fill current needs, a move that's been driven by rapid dropping prices in the Gigabit Ethernet space.
These lowered prices led to double-digit increases in Gigabit Ethernet shipments and an almost 300 percent increase in 10-Gigabit Ethernet shipments in the fourth quarter of 2004, according In-Stat/MDR research. These numbers shows the market shifting from Fast Ethernet to Gigabit Ethernet and 10-Gigabit Ethernet at a time when the overall Ethernet switch market grew only moderately.
In Search of the Killer App
The respondents to the In-Stat/MDR survey echoed what is widely heard in the enterprise space: that without a so-called "killer app" — an compelling application that totally replies on Gigabit Ethernet technology &3151; there just isn't a compelling reason to adopt it immediately. Most desktop users just aren't set up to make use of Gigabit Ethernet technology — much less 10-Gigabit Ethernet technology — which has left the technology in the server room.
One potential killer app: clustered computing, a technology that really is still in its infancy. With clustered computing, a set of computers are treated as one computing whole, with tasks broken down among the machines. In theory, this should enable greatly enhanced performance for computing-intensive tasks and allow cheap workstations and servers to reach supercomputer-like performance. And while clustering technology has come a long way in the last few years, technology like 10-Gigabit Ethernet could enhance computing speed by allowing faster communications between cluster nodes. In this respect 10-Gigabit Ethernet is fighting it out with InfiniBand for high-speed communication.
Another potential killer app: networked storage, where Fibre Channel is widely used. While most technology experts don't consider 10-Gigabit Ethernet to be the perfect solution in a storage situation, it does have the advantage of being able to be deployed in other situations, giving enterprises a way to standardize on a single technology instead of using different technologies for different tasks.
Pricing and Standards Are Key
Still, chances are good that price will continue to drive the market, as well as ease of adoption. One factor: a new 10-Gigabit Ethernet specification, 10GBase-CX4, allowing for 10-GigE data to be transmitted over or four twin-axial copper cable pairs, the same as Fibre Channel and InfiniBand. It's still fairly new, but it's been ratified by the IEEE. And while it's not perfect (it doesn't run over twisted-pair cabling and it doesn't support long-distance throughput) it still would be cheaper to implement than optical networking. But the final roadblock to adoption may be when the IEEE completes a 10GBase-T standard, which could be used over twisted-pair copper cables — a standard that is still several years down the road.