Dumb Switches: A Prediction For 2005
We should expect to see a worldwide shift in what drives enterprise networking technology, at least in terms of speeds and feeds.
January 1, 2005
There's been a spate of 10 Gigabit Ethernet (10GigE) announcements lately. In November, HP announced 24- and 48-port stackable Gigabit Ethernet switches with 10Gbit/sec uplinks. Alcatel followed suit with similar switches and added Power over Ethernet (PoE) capability as a nice twist. A month later, HP announced a stackable 10GigE switch that can act as an aggregator for the gigabit edge switches it previously announced. That got me thinking about why an enterprise might want to drive Gigabit Ethernet to the desktop.
Force10 Networks also showed up in my office recently to tout its high-performance switches. Force10 was once a groundbreaker in 10GigE, so I was curious to hear where its switches were finding a fit. The Force10 folks of course mentioned grid computing and high-performance applications such as those used in the oil industry for analyzing seismic data. No surprises there. One other application mentioned was aggregating residential broadband in Asia. This I found more interesting.
That's because broadband in Asia--particularly in Japan and South Korea--doesn't mean the same thing it means here. There, you get 100Mbit/sec Ethernet service for about $40 a month. Makes our typical 1Mbit/sec DSL service look downright silly, doesn't it? But it also points the way for what's likely to happen in this country. Sooner or later, Fiber To The Curb or Residence (FTTC/R) will enable the sort of bandwidth delivery here that parts of Asia already enjoy.
As a result, we should expect to see a worldwide shift in what drives enterprise networking technology, at least in terms of speeds and feeds. In reality, there aren't many enterprise applications for gigabit to the desktop. I can think of medical imaging, engineering and CAD, and transferring the occasionally oversized spreadsheet or PowerPoint presentation--that's about it. The point is these are the usual bursty applications we've always seen on networks. Sure, more bandwidth makes it easier for us to send a 10Mbyte PowerPoint presentation to the boss when all she really wanted was one slide from it. But that's not enough to drive gigabit to the desktop--that is, unless the technology is dirt cheap. And the only way it's going to get dirt cheap is if the consumer market demands it.
On the consumer side, there are lots of applications that could use all that bandwidth. High-definition interactive movies, sports, and gaming readily come to mind. Combine such a service with your nifty TiVo-like DVR, and you've got something that people will pay for. Sports fans will love the ability to pause action and review other camera angles for key plays (and of course, let's not get into the porn industry).Given the potential size of the consumer market compared to the enterprise market, smart networking vendors will seek to capitalize on these consumer trends. The consumer market needs networking gear where decisions such as permissions and class of service are set at the very edge of the network. The fabric of this network also has to be blindingly fast, with just enough smarts to honor the traffic shaping and policy decisions made by the edge devices.
There's no doubt that this sort of network will evolve. The question is, which enterprise gear providers are going to take advantage of it? Clearly, Cisco Systems is in a good position because it makes both edge gear and blindingly fast, but fairly stupid core devices. Alcatel could also become a force to reckon with, as it parleys its carrier networking experience into enterprise gear.
The third winner may be HP, which is now number two in the Ethernet switching market, behind Cisco. In about six months, expect to see an announcement from HP of just such a fast, dumb switch made for the enterprise. HP has already licensed technology from Riverstone Networks, which makes carrier-class 10GigE gear. So while this current round of 10GigE announcements is interesting, it's my guess that the truly groundbreaking announcements will come later this year.
Editor-in-Chief Art Wittmann can be reached at [email protected].
You May Also Like