The billion-dollar switch

Minor correction: the general press is always confusing bits and bytes, and this post and the Forbes article hopelessly confuse terabits and terabytes. The Nexus 7000 can transfer data at 15 terabits per second, which is fast, but not quite fast enough to transfer Netflix's entire library in forty seconds. By my calculations, that data rate would transfer about 9,000 DVDs in that time, but Netflix says its collection has 90,000 titles.

But apparently the 15Tbit/sec throughput is based on 10Gbit connectivity; the box scales beyond this when 40Gbit and 100Gbit connectivity becomes available.

The 15Tb/Second number is based on the switch fabric capacity, regardless of interface type- although I completely concur that 40 and 100GbE interfces will be able to take better advantage of that type of capacity- allowing the system to enjoy a very long lifecycle.

Agree there is a decimal lost in the calculation somewhere as its 384 seconds, not 38.4. i.e. in the length of time it took me to get logged in to TypeKey the DB could be sync'd and replicated... :)

One other nit while fact-checking- we invest $1B annually into Data Center R&D, the Nexus 7000 is about $250M in R&D spend.

Thanks for the corrections and clarifications.

This type of development is not new and is not a validation of your hypothesis that data centres are the new black.

What this represents is using less _physical_ units to do the same virtual thing. This switch is a progression in existing asset technology, because it combines (at least) two existing technologies into one box (namely, ethernet switching and storage networks).

The Nexus is not much faster than competitive products from Foundry, Juniper et al. And indeed, the speeds are mostly irrelevant since the data centers don't need and cannot generate traffic at this volume.

The speed is important only because a bigger number makes it a more convincing purchase. Note: the top speed will not be possible for four or five years when the modules and line cards that go into it will become available. For example, the C6500 chassis has been available since 1999 but only reached it full capacity in 2005.

I need this box in my networks, so I can have more features in less physical equipment so that I can spend more time working instead of purchasing and deploying.

Well this is a big switch. And Cisco is a great company. But gee it doesn't seem all that big a deal to me. I work for a tiny ( compared to Cisco) company that makes Linux-based switchblades in ATCA chassis, mostly for the telcom market. Trust me, we didn't spend $1b, or $250m, or even $25m to develop our switches. Our largest switch does 48 ports of 1 gig ethernet data ( like phone traffic) and 24 ports of control (used for management functions required in HA environments), and costs less than $10k. It's expected that customers buy them in pairs and have multiple racks, so it wouldn't be all that uncommon to aggregate the same capacity as this Cisco switch using our switches ( or our competitor's). As has been pointed out above, few locations need to switch this much traffic.

Also nobody gets 1 gigabit of traffic through a 1 gig port. 50-80% is more realistic. So of course comparisons about all of Netflix' movies are pretty silly to network engineers. Thats a bit like saying we could run twice as much car traffic by doubling the freeway lanes, when the freeway entrances/exits and feeder routes actually dictate the throughput.

Greg brings up some interesting points, and I just posted to our DC blog about the 15Tb number (http://blogs.cisco.com/datacenter/2008/02/why_youll_want_this_switch_par.html).

Our goal with Nexus was never to have the success of the switch hinge on the bigger/faster argument, although, for now, it is. We believe there are a lot of folks out there like Greg who want to spend more time working and less time purchasing and deploying. Our belief is that once folks have the oppty to dig deeper and check out NX-OS or play with the switch virtualization available through VDCs, or some of the manageability features such as lights-out management, they will see these as the compelling reason to make the investment.

Nexus is certainly not a commodity hardware, at least not yet. Though Cisco's entry in the supply side of cloud computing would certainly attract other players and that would further strengthen the trend of cloud computing (http://cloudcomputing.blogspot.com/)

Thanks for signing in, . Now you can comment. (sign out)

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

URL:
Remember me? YesNo

Comments:

Rough Type is:

Written and published by
Nicholas Carr

Designed by

JavaScript must be enabled to display this email address.

What?