Self aware networks?
I'd be careful of those if I were you.
The switching and server markets are loosely coupled, just like the machines themselves in the data center. When one goes up or down, the other tends to either lead or follow, depending on the technology transitions underway in both markets at any given time. The server market is in a bit of a slump, and the Ethernet switch …
"Whatever's dampening Ethernet switch sales is in full effect in the United States, China, India, and Western Europe. Latin America, Japan, Central Europe and Eastern Europe had single-digit Ethernet switch revenue bumps, but sales in Asia-Pacific (not counting Japan) were down 3.5 per cent year-on-year; in the United States revenue slipped 6.4 per cent and in Western Europe it was even worse with a 10 per cent decline."
Whatever's dampening? I guess the article author has not experienced the global financial melt-down of late. Hate to break it to them, but all sales are down save for doomsday bunker supplies. Those are booming.
As a buyer of switch gear in a data centre, i can attest to the "we like 10gig but we don't like 10gig prices" view.
A clumsy approach at the minute is teaming multiple gig connections and the switching is still the bottleneck in a lot of our vm arrays.
95% of my infrastructure is 1g switches. we'd love to upgrade the old distro / access layer 6500's but the capital outlay at the moment is crippling. Can't fault Cisco kit, but would like to see a bigger share for Juniper or any of the other minnows feeding the scraps. Hopefully this would lead to a bit of competition and a better deal for tech budgets.
Couldn't agree more. Looking at replacing a lot of our core shortly as we're ditching our trusty FC fabic for a converged iSCSI ethernet network - and quite honestly it's a nightmare. Mainly a ProCurve and Juniper shop (although use Cisco 6500 too) but trying to get above 1Gig for anything other than the backbone is far too costly. I hate planning a new topology that's designed to last a good 5 years + and on day one going to have to use LACP / NIC teaming just to get the baseline performance required!
1-Gig would be great outside of inter-switch backbone links but just far to expensive still.
The Juniper and Netgear stories are very appealing... until the price tag comes. The HP story isn't even appealing before the higher-than-the-competition price. At least with Cisco, you knew you were going to get reamed anyway. So yes, 10gE adoption won't heat up until ASIC makers bring the cost down significantly, or datacenters will just make do with bonded gE and leave it at that for now. You can significantly cut down on cable clutter with rack-switches that have extra 1gE or 1-2 10gE backbone links, at least, rather than the old "run everything to the core switch" paradigm.
Too bad a pair of 10gE backbone links more than triped the price for the juniper EX switches. They're so perfect in all other respects.
I deployed a Brocade 10Gbe solution in our datacentre this Year, mainly for VM consumption - works a treat and good value when compared to the previous Cisco 1Gbe kit. The only time the bandwidth 'tops out' is during overnight backups (which now take considerably less time of course!)
Merry Christmas one and all!
Hell, I'd be happy if they replaced that last 100M switch up on the executive floor.
Yeah, don't ask me how, but the rest of us are on faster switches and I always find myself wondering why the it takes so damn long for simple thing to finish when I'm working up there. But maybe they'll get around to next year at the same time they're replacing our 1970s phone system with a shiny new VoIP one.
Biting the hand that feeds IT © 1998–2021