Forgot your password?
typodupeerror
The Internet Technology

Cisco Introduces a 322 Tbit/sec. Router 281

Posted by kdawson
from the one-loc-per-second dept.
CWmike writes "Today Cisco Systems introduced its next-generation Internet core router, the CRS-3, with about three times the capacity of its current platform. 'The Internet will scale faster than any of us anticipate,' Cisco's John Chambers said while announcing the product. At full scale, the CRS-3 has a capacity of 322Tbit/sec., roughly three times that of the CRS-1, introduced in 2004. It also has more than 12 times the capacity of its nearest competitor, Chambers said. The CRS-3 will help the Internet evolve from a messaging to an entertainment and media platform, with video emerging as the 'killer app,' Chambers said. Using a CRS-3, every person in China, which has a population just over 1.3 billion, could participate in a video phone call at the same time. (Or you could pump nearly one Library of Congress per second through the device, or give everyone in San Fransisco a 1Gbps internet connection.) AT&T said it has been using the CRS-3 to test 100Gbit/sec. data links in tests on a commercial fiber route in Florida and Louisiana."
This discussion has been archived. No new comments can be posted.

Cisco Introduces a 322 Tbit/sec. Router

Comments Filter:
  • by NevarMore (248971) on Tuesday March 09, 2010 @03:48PM (#31417942) Homepage Journal

    Kidding, but you know someone is going to seriously ask that sometime today.

  • by ravenspear (756059) on Tuesday March 09, 2010 @03:48PM (#31417948)
    The new standard in router benchmarks for the 21st century!
  • by Harik (4023) <Harik@chaos.ao.net> on Tuesday March 09, 2010 @03:48PM (#31417952)

    If the first poster doesn't have a comment like "Yeah I'm using one of them right now, my internet is blazingly fast", it's a wasted opportunity.

  • by Jeremy Erwin (2054) on Tuesday March 09, 2010 @03:50PM (#31417990) Journal

    MSRP starts at $90,000. source [cisco.com]

    • by Hadlock (143607)

      Great! I'd been eyeing the CRS-1 for a while, but now that the CRS-3 is out, the price on the CRS-1 will finally drop down enough that I can complete my beowulf cluster of failed linux PDAs. Looks like CRS-1s are going for $20,000 on ebay used.

    • by mcrbids (148650) on Tuesday March 09, 2010 @04:01PM (#31418178) Journal

      Strangely, at $90,000 a pop, this strikes me as rather cheap. I wonder if that's a "rate limited" model so that you have to pay big bux more in order to get the full capacity?

      • by Wesley Felter (138342) <wesley@felter.org> on Tuesday March 09, 2010 @04:04PM (#31418216) Homepage

        Strangely, at $90,000 a pop, this strikes me as rather cheap. I wonder if that's a "rate limited" model so that you have to pay big bux more in order to get the full capacity?

        You wish. For $90K you probably get an empty chassis... the smallest available empty chassis, that is.

        • Re: (Score:3, Informative)

          by afidel (530433)
          Hmm, a 6509E chassis is only $9,500 list and can switch at 720Gbps (when equipped with Sup720). Of course by the time you add two Sup720's with 3BXL forwarding engines you're up the $63,000 list and you have nothing but four empty 10GBps slots.
        • by billstewart (78916) on Tuesday March 09, 2010 @07:57PM (#31421170) Journal

          Based on a brief look at Cisco.com [cisco.com], it looks like the CRS-3 scales from a single 4-slot chassis up to an 1192-slot multiple-rack array, so the amount of backplane capacity you get depends on what size chassis and how many chassis you want to chain together, as well as what flavors of interface cards you put in them. (A lot of the processing capacity is on the cards, which is how you get things to scale to carrier-class.) The small box is going to have supervisor CPUs and probably control-plane, and you'll presumably want redundant power supplies of some sort (though that may be DC if you're in a carrier environment), and probably a couple of GigE interfaces on the supervisor card, but it's not the kind of platform you buy without buying some hefty interface cards, which is where most of your money'd be going.

      • by amn108 (1231606)

        In the kind of world we live in (which "we make ourselves") it wouldn't surprise me if one had to pay per unit of bandwidth transferred by the router. :|

      • Re: (Score:2, Insightful)

        by hamisht (197412)
        when I first read your comment I thought you said you would have to "pay for the bug fix" to get full capacity, even after re-reading I think my initial parsing still makes sense...
    • by Sponge Bath (413667) on Tuesday March 09, 2010 @04:05PM (#31418234)
      I take it that $90K is for an empty shell and you must buy plug-in modules to actually accomplish anything.
      • by DarkOx (621550)

        Yea and I can only imagine the SMARTNET costs...You think TAC will call you back in less than two hours if you own one of these things.

    • by Aqualung812 (959532) on Tuesday March 09, 2010 @06:16PM (#31420040)
      Actually, the question on my mind is if this device is really going to be used to just route bits at layer 3, or if such massive hardware is going to sell more as a very fast deep packet inspection layer 7 device. I think there are ISPs like AT&T that would love to go deeper...in every way.
      • Re: (Score:3, Informative)

        by afidel (530433)
        Deep inspection is done at the edges.
      • Re: (Score:3, Interesting)

        by Bender0x7D1 (536254)

        or if such massive hardware is going to sell more as a very fast deep packet inspection layer 7 device.

        There is no way they would be able to do deep packet inspection at those kinds of speeds. Just think about a 1TB hard drive. Now imagine 300 of them. Now, you want to inspect all that data in 1 second. It's just not going to happen. That's why it's listed as a core router - it's job is to move a LOT of data as fast as possible. In fact, other routers do extra work to reduce the processing done by t

  • by Archaemic (1546639) on Tuesday March 09, 2010 @03:55PM (#31418080)

    I'd make a joke about how the internet can now handle the flow of porn through it, but I'm sure that with one of these routers, I've already been beaten to the punch!

  • jaded, who care? (Score:5, Insightful)

    by Neil Watson (60859) on Tuesday March 09, 2010 @03:56PM (#31418094) Homepage

    Between terrible last mile infrastructure and ISP throttling I can't help but sarcastically comment big freaking deal.

    • by kenp2002 (545495) on Tuesday March 09, 2010 @04:11PM (#31418306) Homepage Journal

      Between terrible last mile infrastructure and ISP throttling I can't help but sarcastically comment big freaking deal.

      We'll they can't complain now that there isn't enough bandwidth so they have to meter it now.

      Cisco as I see it has a vested interested in ensuring that the net remains neutral to push these kind of product upgrades. Coupled with premise end-point equipment it stands that they would want more bandwidth use and leverage monitoring, rather then metering, Internet use.

      Metering is a waste, monitoring and then selling said info, there is where the money will be...

      • Actually, I'm pretty sure they sell monitoring hardware as well (can't remember which story it was linked to but I think it was Comcast).

        They win either way.

    • I suppose if your ISP got one of these you might find an improvement - especially if Google opens shop next door and wants to offer you a Gigabit connection to your house, your ISP might jump up to stay in competition.

      And, just in case you weren't aware, there are cases where networking exists outside of the internet. True story!

    • by OverlordQ (264228)

      jaded, who care?

      The backbones?

    • Re: (Score:3, Interesting)

      by timeOday (582209)
      You're so bothered by the problem you don't even care about the solution?
  • I can watch TV... On the Internet!

    John Chambers: Man of Vision!

     

  • CIA/NSA software loaded to do deep packet inspection?

    -Hack

    • by gknoy (899301)

      Wouldn't a more efficient way to do that be to just route ALL the traffic to a separate machine (or set of machines) to do the deep packet inspection?

      • Yes. That's essentially what they did with AT&T when they were doing some big-time packet snooping. They spliced the fiber and ran it to it's own floor where it was analyzed by a separate system.

  • Ars had a story yesterday about Cisco: Cisco: Internet to change forever Tuesday (place your bets!) [arstechnica.com]

    Is this the thing that will change the internet??
    • Probably. 322 Tbit/sec is quite a lot.

      • by Albanach (527650)

        Probably. 322 Tbit/sec is quite a lot.

        Well, it's 3.5x faster than their fastest CRS-1 that was available yesterday. So it's an improvement, but not exactly a revolution.

  • Geek Porn (Score:4, Insightful)

    by keithpreston (865880) on Tuesday March 09, 2010 @04:04PM (#31418222)
    Ok, 322Tbit/sec is cool and all, but where is the geek porn of it? Images, technical details and specifications? Otherwise it is vaporware to me.
  • Cables? (Score:4, Interesting)

    by kyrre (197103) on Tuesday March 09, 2010 @04:04PM (#31418230)

    What kind of wire would this router need? Is a single fibre cable enough for this kind of bandwidth? What is the limit of a fibre cable?

    • Re:Cables? (Score:5, Funny)

      by Kagura (843695) on Tuesday March 09, 2010 @04:12PM (#31418332)

      What kind of wire would this router need? Is a single fibre cable enough for this kind of bandwidth? What is the limit of a fibre cable?

      Eleven.

    • Re: (Score:3, Informative)

      by TheSync (5291)

      What is the limit of a fibre cable?

      Alcatel-Lucent demonstrated 25.6 Terabit/s [prnewswire.com] in 2007 using 160 Wavelength-Division Multiplexed channels of 160 Gbps each.

  • by bughunter (10093) <bughunter&earthlink,net> on Tuesday March 09, 2010 @04:08PM (#31418270) Journal

    So, I've been waiting for something better than 150 kB/s service for years, despite the promises by AT&T and Verizon that they're "rolling out" fiber to the home. Not my home.

    When can I finally stream in real time at least one channel of video content that's not so compressed that it's unwatchable? At a subscription rate of under $40/month? When that happens, I'll be impressed.

    However, I'm fearing that USians have been living under monopoly conditions of artificial bandwidth scarcity for so long that we're going to let the AT&Ts and Verizons charge us an arm and a leg for this kind of service in the near term.

    • by olden (772043) on Tuesday March 09, 2010 @04:34PM (#31418638)
      Amen to that.
      I live in Palo Alto, heart of the Silicon Valley I was told. Fastest connection I can get (without having to take a 2nd mortgage, that is): 768 kbit/s. And, with a static IP, the same price as 9 years ago. WTF?!?
      In the meantime, French ISPs are addressing complaints that 22 Mbit/s VDSL is a bit old-school by offering 100 Mbit/s FTTH [www.free.fr] (phone and TV included, of course), Japanese get Gigabit for ~60$/mo [auone-net.jp]...
      AT&T, I'm glad you're upgrading your equipment at long last... Now when can I get better than 3rd-world connection?
    • by b0bby (201198)

      Hmm, I got FIOS this year, it's about $40/month for the internet portion, and I can stream HD Netflix movies which look great with no problem. So I'd say find out where Google is rolling out their fast fiber & move there ;)

    • by jadin (65295)

      despite the promises by AT&T and Verizon that they're "rolling out" fiber to the home. Not my home.

      Every local ad I see has an asterisk next to it explaining that the fiber stops at the last mile to my home. Makes me wonder how many people are signing up thinking they are getting fiber connections..

  • to fap furiously. Do want.
  • by wonkavader (605434) on Tuesday March 09, 2010 @04:14PM (#31418372)

    "At full scale, the CRS-3 has a capacity of 322Tbit/sec., roughly three times that of the CRS-1, which was introduced in 2004."

    That was six years ago and we're only tripling the speed? Is it cheaper? Smaller?

    Moore's law (which doesn't work in every way, but it certainly works for the computing processors in this thing) would suggest that this thing has a lot more CPU power than the CRS-1. (In six years we'd expect somewhere between 8 and 32 times the oomph.) And yet they only encumbered it with three times the bandwidth.

    I'm worried that a lot more processor power is going into filtering. Cisco is one of the big anti-network neutrality advocates. They want to sell the machines to impose the rules.

    If this machine isn't lower power or smaller or cheaper or just built incompetently, then the real story here isn't it's bandwidth -- it's its power for adjusting traffic for increased profits.

    • by ishobo (160209) on Tuesday March 09, 2010 @04:24PM (#31418524)

      Moore's law is about transistor density, not computing power.

    • Re: (Score:2, Insightful)

      Moore's law (which doesn't work in every way, but it certainly works for the computing processors in this thing) would suggest that this thing has a lot more CPU power than the CRS-1. (In six years we'd expect somewhere between 8 and 32 times the oomph.) And yet they only encumbered it with three times the bandwidth.

      Moore's law [wikipedia.org] applies to the number of transistors on an integrated circuit and has absolutely nothing to do with bandwidth. Chip throughput is much more a function of the chip architecture than the number of transistors on chip. Even if chip throughput was somehow correlated to Moore's law, there are still unrelated inefficiencies in the physical layer that are very complex and difficult to overcome.

  • I'm sorry but my CRS-3 (can't remember shit) Syndrome is running quite fast today. It's currently deleting the question before you even ask it and creating a space/time continium loop meaning we'll have to repeat this day forever

  • Imagine how much traffic could be routed to collection clusters on behalf of your favourite three letter agency.

  • It runs QNX (Score:5, Interesting)

    by Animats (122034) on Tuesday March 09, 2010 @04:22PM (#31418484) Homepage

    Like all Cisco high-end routers, it runs QNX Neutrino. The version used in these routers has a 12KB (not MB) microkernel. Almost all the packet handling is in FPGAs, but the supervision, error handling, etc. are in Cisco applications running on QNX Neutrino.

    • Re: (Score:2, Interesting)

      by Anonymous Coward

      yah QNX is really nice, drivers are running in used space so you can queue up redundant drivers against hardware, should a driver crash the driver in waiting can start up and take over even before the next packet comes. Kernel level instrumentation, no need to restart OS to restart drivers, lots of benefits. Anyway Ciscos have their own version of QNX 6 they have tailored for themselves..

  • by GPLDAN (732269) on Tuesday March 09, 2010 @04:26PM (#31418542)
    In Japan, it's pretty easy even in rural areas like Kyoto to order a 100Mb connection and get it at a reasonable rate.

    In the States, we're playing on DSL lines that have 2Mb down, when they train up right (which they only do maybe 50% of the time) and other people are using Cable (Charter, Comlast, etc) and maybe that is 5 or maybe 10Mb down. If you are very lucky (and have the coin) maybe you are on AT&T uVerse or Verizon FIOS, and they could give you 100Mb, but you'd pay through the nose for it, and it would be asymmetrical. Most likely (the UVerse people I know) you are getting 10 down.

    Now here comes Johnny Chambers saying this beast in the core could give GIG (1000Mb/s) to every person in San Francisco. Johnny's comb over is going to his brain. Just because a TR2N sized CRS-2 with enough horsepower to make the TRON MCP break down and cry comes into the provider core doesn't mean SHIT to you, the end user. Here in the states we won't see Japanese style connectivity for another 10 years. We're being left in the fucking stone age, because they money isn't there to build out past the core.

    It pisses me off when Johnny tries to hype and pimp that stock price up, and they use multi-threading and distributed fabrics to get that speed, but we all know it's moving at snail's pace, the industry is consolidating, and unless you live where fiber is, forget it. And save me the "USA is so much bigger than Japan" argument, too. We don't see these speeds in our major cities, like NYC or Atlanta, SF or Chicago. Nothing even close. the SONET rings in these cities are still selling OC multiples at insane prices. It's still fucking 1996 in America.
  • Cisco built a bigger, faster router.
    Nothing to see here, folks. Move along. Move along. Just Cisco marketing engaging their HYPErdrive by claiming to "...change the Internet forever..." and other HYPErbolic phraseology.

    Please...
  • Bandwidth cap (Score:5, Insightful)

    by vlm (69642) on Tuesday March 09, 2010 @04:42PM (#31418788)

    Using a CRS-3, every person in China, which has a population just over 1.3 billion, could participate in a video phone call at the same time. (Or you could pump nearly one Library of Congress per second through the device, or give everyone in San Fransisco a 1Gbps internet connection.)

    Or, could exceed their monthly bandwidth "cap" in 155 microseconds. So, what good is it?

  • It's a subscription model. You pay them 90k/quarter and they keep bringing you their latest vaperware. Nice!

  • Using a CRS-3, every person in China, which has a population just over 1.3 billion, could participate in a video phone call at the same time. (Or you could pump nearly one Library of Congress per second through the device, or give everyone in San Fransisco [their own private] 1Gbps internet connection.)

    So that means we'll need maybe a thousand of these things to pipe the whole world's bandwidth? Doesn't seem like much of a market.

  • A very slight rewording:

    'The Internet will scale faster than any of us anticipate,' anticipates Cisco's John Chambers.

  • Dr. Chris Centeno posts several times at the end of the article and addresses most of the issues raised here. Definitely worth reading.

  • Or reduce network reliability by reducing redundancy and introducing more critical choke points.

  • by Glock27 (446276) on Tuesday March 09, 2010 @09:07PM (#31421766)

    or give everyone in San Fransisco a 1Gbps internet connection

    Or give everyone in San Francisco a 1 Gbps Internet connection! :-)

HEAD CRASH!! FILES LOST!! Details at 11.

Working...