• United States

10G: Worth your attention

Aug 18, 20033 mins

Dismissing 10G Ethernet for the enterprise comes easily to many industry watchers. “That’s just too much bandwidth for your average user organization,” they typically scoff. “Truly,” they say, “10G belongs in the carrier domain.” Even the network engineer I asked to write a story about 10G switches for the enterprise argued that point at first. The gist of his argument: All the action is in the public network.

Maybe, maybe not. 10G Ethernet is a huge amount of bandwidth, and carriers with their chunky cross-country and metropolitan-area routes have good reason to grab up 10G switches. For them, 10G makes for a reasonable, increasingly cost-effective alternative to traditional WAN technologies such as packet over SONET and ATM.

But a growing number of enterprise users are finding 10G a compelling story, too.

Today, 10G, maturing nicely since stamped with the IEEE’s approval in mid 2002, is worth everybody’s attention. The reasons are piling up, as we explore in our special “10G: Big, bad bandwidth” coverage.

For starters, 1G oftentimes isn’t quite enough anymore for many corporations. The 10G version of the technology is proving useful on high-volume storage-area networks and back-up segments, within data centers heavy with 1G-based switching and routing cores, and for carrying bandwidth-intense applications such as IP telephony and real-time audio and video multicasting.

And then comes the basic maturation of a technology that pushes it from the “needs to be watched” to the “needs to be sampled” stage. In this case, technology refinements and cost reductions are coming together to make 10G, at least in trial mode, a palatable decision.

Take just one vendor, Extreme Networks, as an example. The company is working out the flakiness in its initial “get to market fast” 10G blades, alleviating the throughput problems that limited interconnections to 4G bit/sec, and the latency and jitter errors that botched transmissions. Its Mariner switch, set for release by year-end, will provide six-port 10G blades at $8,000 apiece.

Finally, service providers are talking about edging into the 10G enterprise action. For example, AT&T has talked about developing 10G metropolitan-area network (MAN) services that would find their way to the table in another year or so. The potential there is grand, as the enterprise LAN merges seamlessly with the MAN and then even the WAN.

Gartner expects to see a big boost in 10G pickup beginning next year. It forecasts that 10G volumes will begin at least tripling year over year, from 5,000 ports sold next year to 185,000 in 2007, with the average per-port pricing dropping to about $7,700. Now those figures are nothing to snigger at.