Stories
Slash Boxes
Comments
Search

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

IEEE Sets Sights on 100G Ethernet

Posted by Zonk on Tue Dec 05, 2006 12:33 PM
from the aint-no-mountain-high-enough dept.
Networking The Internet Technology IT
coondoggie writes to mention a Network World article about the IEEE's new 100G Ethernet initiative. The organizing body's High Speed Study Group has voted to try for the 100G standard over other ideas, like 40Gbps ethernet. From the article: "The IEEE will work to standardize 100G Ethernet over distances as far as 6 miles over single-mode fiber optic cabling and 328 feet over multimode fiber. With the approval to move to 100G Ethernet, the next step is to form a 100G Ethernet Task Force to study how to achieve a standard that is technically feasible and economically viable, says John D'Ambrosia, chair of the IEEE HSSG, and scientist of components technology at Force10 Networks." With video download services and interactive media becoming ever more the focus of internet startups, the organization is eager to offer a way to aggregate pipes in the coming years. The current thinking is that achieving these speeds will be reached by advancing bonding techniques for 10G signals over multiple fibers.
+ -
story
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

IEEE Sets Sights on 100G Ethernet 50 Comments More /

 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • As it is, your average desktop will not handle anything even close to 100G Ethernet. At that point, your bottleneck is the PCI or PCI-X bus. As the bus has been one of the slowest PC components to innovate, I see these new, ultra-high speed Ethernet standards as only benefiting backbone providers, etc., for many years to come.

    • Re:What's in it for desktop users? (Score:5, Insightful)

      by corsec67 (627446) on Tuesday December 05 2006, @12:45PM (#17115486) Homepage Journal
      What, you think that any ISP would actually allow downloads fast enough to use over 100baseT?

      Really, even full 10baseT (as an obtainable download speed, not just the home->CO link speed) would be an improvement to many people.
      • They would allow it if it was cost effective, some countries already have 100mbit to the home. To get that requires a huge backbone to start with and it needs to be available to the Telcos/ISPs at reasonable prices.

        Bonding huge links together will be quite a feat, as far as I know the main bonding protocols in use now (etherchannel, LACP, etc) are based on current ethernet standards so may need some reworking, unless the large links are already using Ethernet (DWDM maybe?). Then there's the small matter of
      • What, you think that any ISP would actually allow downloads fast enough to use over 100baseT?

        Believe it or not, some people use LANs for things other than accessing the internet... The internet connection speed becomes unimportant if the network is actually a SAN.

        Really, even full 10baseT (as an obtainable download speed, not just the home->CO link speed) would be an improvement to many people.

        We're reaching the point now where I've stopped caring so much about download speed (I have an 8Mbps DSL) - upload speed is becoming a serious headache since on most ADSL lines (at least in the UK) it tops out at ~340Kbps. At that upload speed you're talking about ~45ms per MTU sized (1500 byte) packet - that's quite a lot of latency jitter and can cause serious problems for realtime applications such as VoIP, which often have jitter buffers of only around 100ms long.

        • - upload speed is becoming a serious headache since on most ADSL lines (at least in the UK) it tops out at ~340Kbps

          Well, apart from the fact that I seem to get 448kbps up (and I'm not alone) you can always try paying a little more.

          For example - these people [lawyersonline.co.uk] are offering "Up to 24,576K download speed", and "Up to 1,331K upload speed" on a residential use basis for only £85.47 per quarter (£28.49 per month) I don't know the VAT status of that quote. Or they also offer this [lawyersonline.co.uk].

      • ADSL2+ speeds long ago left 10BaseT speeds behind. Now that ADSL2+ is becoming obsolete, and fiber is going in, expect that 3rd world countries will pick up the kit for cheap. 20Mbps sustained downloads are pretty common in Europe, 8-10Mbps in the UK. I know of many companies that have a single ADSL line for the 20 to 50 PCs in the office, its enough bandwidth for casual internet use.

        Of course, if you live in a country with a corrupt administration and a broken telecoms regulator, then you will never know t

    • Re:What's in it for desktop users? (Score:5, Funny)

      by UnknowingFool (672806) on Tuesday December 05 2006, @12:49PM (#17115538)
      Well that swimsuit supermodel in the magazine I've been lusting after will never date me, but I can dream about it can't I? At least 100G Ethernet to the desktop might be realized in my lifetime. A supermodel, not so much. :P
      • But w/ a 100Gb Ethernet you can download the Real Doll data of the supermodel to your desktop fabber real fast!
      • At some point in your lifetime, 100G Ethernet will be available on the market, and you will be willing to pay for it. Would you ever be willing to pay for a supermodel, which has been on the market since known human history?

        See, that's the problem.
    • Well, yes, that's what's in it for desktop users; networks upstream can scale further.

      And who says you have to connect it to PCI-(E|X)? Hook it up directly to a HyperTransport link and talk to other systems on the network at reasonable speeds.

      • Re:What's in it for desktop users? (Score:5, Interesting)

        by Znork (31774) on Tuesday December 05 2006, @01:18PM (#17115910)
        Heck, with 100Gb ethernet, who says you have to _have_ PCIe; once you reach those speeds it would be entirely viable to move most PC components to their own ethernet bus/network. Imagine having your NVidia graphics units connected to your LAN and usable from any of your PC's; plugging another unit into the network makes it instantly accessible by all devices as tho it was more or less local hardware. Etc.

        SAN is storage moving that way, we might very well expect other components to move in the same direction.

        Of course, expect a horde of crap patent applications for shit like 'graphics acceleration _over a network_' just because the technology becomes feasible. Which may drive prices through the roof and/or hold development back a decade or five.
    • Sure, your average desktop will not handle 100G Ethernet, but what kind of content or traffic could possibly require that much desktop bandwidth? And since this is over fiber, there are very few desktop networks out there you could even plug the PC into.

      This will certainly be a backbone technology, and a server technology. But this particual technology doesn't seem likely as a desktop technology in the near future.

    • Re: (Score:3, Informative)

      by jellomizer (103300)
      Well a faster Pipe for the ISP allow them to increase max speed for their users. So say they had a 100GB Eathernet for their backbone internet connection. That means they could possible increase the max speed for their customers to closer to 10MB as of right now most U.S. ISP tend to cap to 5MBS Anything above that could be to much demmand on their systems. So you as a desktop can see an impovement.
    • Right now actually the PCI bus can't even take advantage of a 10G Ethernet card - you'll see that in real world conditions standard server class hardware is pretty much capped at 2G because of bus limitations.
    • Because it's not about the desktop, but yes, desktop users benefit.

      As backbone providers get more capacity, they can deliver faster speeds to their customers (your ISP and the ISPs of the websites you visit.) Ethernet is a very cost-effective physical transport layer as it removes some of the administration headaches involved with point-to-point links. This will eventually drive down the cost of fast backbones, allowing more bandwidth for less money.

      Your average desktop user doesn't have a need for 10G, muc
    • >>>"As it is, your average desktop will not handle anything even close to 100G Ethernet"

      Key words there "As it is"; If they build it, etc...

      Plus, I'd quite happily have 100G to the house. It would not be for one computer, but for the four that I currently have, plus who knows how many I'd have by the time they roll it out.

      A couple of apps I might use it for [pipedream]:
      thin client gaming to Google-games(TM), where the googleserver does all the game crunching, HDR etc. I might want to Slin
    • You are right. Here are some numbers for the curious, nothing comes close to 100 Gbit/s:

      PCIe x16 (2.5 Gbit/s per lane, 8B/10B encoding): 32.0 Gbit/s bidirectional (64.0 Gbit/s of aggregated bandwidth)
      PCIe x8 (2.5 Gbit/s per lane, 8B/10B encoding): 16.0 Gbit/s bidirectional (32.0 Gbit/s of aggregated bandwidth)
      PCIe x4 (2.5 Gbit/s per lane, 8B/10B encoding): 8.0 Gbit/s bidirectional (16.0 Gbit/s of aggregated bandwidth)
      PCIe x1 (2.5 Gbit/s per lane, 8B/10B encoding): 2.0 Gbit/s bidirectional (4.0 Gbit/s of agg
      • You mean like Myricom

        http://www.myri.com/Myri-10G/10gbe_solutions.html [myri.com]

        Not bad prices, list is 795USD for a fibre optic card and 500USD for a SR optic or 900USD for a LR optic. A CX4 card is only 695USD. With switches like the HP Procurve 2900 having 10GbE CX4 as standard, I predict that 2007 is the year when 10GbE really moves mainstream.
    • And some of us would be happy if they only marginally bumped up the capacity, but worked on the latency issues instead. Something with Gigabit to 10Gigabit speed, but with less than 10 usec of latency would be a good start. Myrinet for everyone, since some of us use Beowulf-type clusters for work, rather than humor.
    • News flash, your average desktop cannot even handle a 1Gbps link, let alone a 10Gbps. Experience tells me that you will see about 30MBps max out of a 1Gbps link on desktop hardware. You need server grade kit to go faster. I can max out a dual bonded 1Gbps link on my servers for example.
    • Even if the desktop user never has direct access to it, he will benefit. This sounds like a great way to solve the last mile problem; a 100-gigabit line will support 500-1000 users at 100 megabits each.

    • At 100Gbps, your processor's L1 cache is a bottleneck.

      • Re: (Score:3, Interesting)

        by GigsVT (208848)
        Don't worry about content. We'll find plenty of things to fill that with.

        Look at a platform like Second Life. It uses a very simplified version of CSG 3d modelling because of bandwidth constraints of current broadband. Now imagine Second Life with fully deformable meshes and high resolution textures in a world that is downloaded faster than you can move through it.

        Anyway, we'll find plenty of things to do with more speed, we always do.
        • I imagined it--and it's still totally meh.

          Perhaps they could add something more fun than "Buying Stuff" instead. Even Pokemon had a game to play while you were mindlessly collecting "them all".

  • I prefer Bill Watkins' take on it. (Score:5, Insightful)

    by Tackhead (54550) on Tuesday December 05 2006, @12:47PM (#17115514)
    "The need for 100G Ethernet is growing as IP video and transaction-intensive Web 2.0 applications are exploding across the Internet. Companies such as YouTube regularly add 10Gbps service pipes to meet growing demand, and carriers will need a better way to aggregate such links, industry watchers say."

    - From TFA.

    Which is all well and good, but for honesty, I prefer Bill Watkins' take on it.

    "Let's face it, we're not changing the world. We're building a product that helps people buy more crap - and watch porn."

    Bill watkins, CEO of Seagate [cnn.com]

  • don't trust such initiatives (Score:5, Insightful)

    by SilentGhost (964190) on Tuesday December 05 2006, @12:47PM (#17115520) Journal
    328 feet - it's a good standard, but I like 100 metres better.
    • Insightful? [google.com]

      While I do prefer the metric system, it would be nice if the mods remembered to have their funny detectors on. =)

        • Re:don't trust such initiatives (Score:4, Informative)

          by slcdb (317433) on Tuesday December 05 2006, @01:27PM (#17116034) Homepage
          The actual standard will state the maximum distance in meters, not in feet. Whoever wrote the article did the conversion from meters to feet.

          Just search IEEE 802.3 for yourself. You'll find no mention of "feet". Everything in there is measured in meters.
    • Interesting. I was wondering where they pulled that spec out of. My next question is: If they based this one off of metric, why did they base the other one off of Old English?

      • Re:don't trust such initiatives (Score:5, Informative)

        by tool462 (677306) on Tuesday December 05 2006, @01:22PM (#17115962)
        Pause for a moment and realize that 6 miles is approximately 10km, which is probably the "real" spec. Not only is it metric, but it's a power of 10, which gives me lots of warm fuzzies.
        • Oh tell me about it.

          I have no idea how people can use imperial measurements. All we have to say is '1 litre'; they have to somehow remember 2.11337641 pints!!

          • Re:don't trust such initiatives (Score:4, Funny)

            by trb (8509) on Tuesday December 05 2006, @02:49PM (#17117538)
            I have no idea how people can use imperial measurements. All we have to say is '1 litre'; they have to somehow remember 2.11337641 pints!!

            And they say 1 pint, and you have to remember 0.47317 litres. Tag, you're it.

            28 g of prevention is worth .454 kg of cure. Ick!

  • Uplink (Score:4, Interesting)

    by GigsVT (208848) on Tuesday December 05 2006, @12:51PM (#17115570) Journal
    What I really want to see is higher uplink ports on SMB hardware.

    Right now, if I want to make a medium size network using lower cost components, it might look something like 5- 24 port, 100-meg switches with 1 GB uplink to a big GB switch.

    The bottleneck here is those uplinks. Each 100meg switch has plenty of backplane, and so does the gigabit switch, but those 100 meg 24 port switches have to share 1GB each to the backbone MDF.

    So I really don't care about PCs or network cards or whatever, just give me 10GB links that I can use between switches without having to pay for overpriced Cisco crap.

    • Oversubscription models have not ever crossed your mind, have they?

      1:2.4 oversubscription isn't bad at all. Do you really think that a 24pt 100mbps switch needs 10 GIGABIT uplinks in order to work well? If so, I'm sure that Extreme or Force10 would love to sell you some hardware.
        • Yes, in which case, if you want pure gigabit performance, YOU HAVE TO PAY FOR IT.

          Can you legitimately justify to me, or anyone else, that an SMB network needs 1gbps access-layer switches, with 10gbps uplinks to distro/core layer switches? If that's the case, then I'll show you a network that needs to be running on something like Cisco, Extreme, or Foundry, and NOT your 'lowpriced SMB switches'.

          You can't build a sports car out of turds and baling wire. Well, you could, but you shouldn't expect much from it.
          • Maybe, but at the moment there is no way on earth we could take 1Gb to the edge at work. The network is two big. Think up to 1400 outlets a patch room, think over a dozen patch rooms in the building. We are not using cheap SMB switches either, it's a combination of managed enterprise HP Procurve and Cisco.

            At the moment it is 100Mbps to the edge, 1Gb uplinks in the patch room, and 1Gb (sometimes two) uplinks out the patch room. We really need 10Gb uplinks out the patch rooms just to get the performance level
            • There are full-rack chassis capable of terminating 1200+ 1gbps links, and can support multiple 10gbps uplinks. Specifically, the Foundry NetIron MLX-32. I will be blunt - if you've got 50 24-port switches in a patch room, I can geniunely state that your network planning skills are atrocious. There *are* better solutions out there. You may not LIKE the cost, but to be honest, unless you're ready to build your own switch, it's always a toss-up between cost and features. If you can't spend to get the features,
            • You're arguing a ridiculous point - that there should never be oversubscription in a network, anywhere.
    • You could use something supporting etherchannel and bond a few 1GB links together. We use that to great success, admittedly with using Cisco kit but there's plenty of other companies around making kit that supports channel bonding.

      Incidently what are your users doing that maxes out gig uplinks? We have 96 ports sharing 2x1gig uplinks all over the office without problem, but none are particularly heavy traffic users.

    • Re:Uplink (Score:4, Insightful)

      by Amouth (879122) on Tuesday December 05 2006, @01:14PM (#17115868)
      just wanted to note.. while yes Cisco is overpriced they most certainly are not crap.. they do exactly what they are supposed to do and they do it well. if you are looking for something that is above the average Joe's network you are going to have to pay whether it be Cisco, foundry, or anyone else - most people that have Cisco switches don't use half the features that they get with them.. they just plug them in and run.. it is their configurability that makes them rock.
    • Channel bonding is your friend. Assuming a top of rack 48-52 port gig switch you can take your 1:48 over subscription down to 1:12 or 1:6. Depending on what your doing a 48+4 switch gets you a nice 1:12 over subscription and 1:24 with a failure assuming you can split up the vlans.

        • Re: (Score:3, Interesting)

          by WuphonsReach (684551)
          If you don't need VLANs or managed switches, try the "smart" switches which fill the niche between completely unmanaged and fully managed switches. Then you can get things like a 48-port gigabit smart switch for around $1500. Which gives you 40 ports for end-users and 8-ports for uplink or backbone use. Even some of the less expensive "smart" switches support VLANs, but you have to configure using a web browser.

          They aren't the fastest things in the world, but at least they do trunking.

          (Heck, I have a

  • Its already done (Score:3, Interesting)

    by warrior_s (881715) <{kindle3} {at} {gmail.com}> on Tuesday December 05 2006, @01:08PM (#17115796) Homepage Journal
    Not exactly but Bell Labs did something like this in March http://www.lucent.com/press/0306/060308.coi.html [lucent.com]

  • *sigh* (Score:5, Insightful)

    by M0b1u5 (569472) on Tuesday December 05 2006, @02:26PM (#17117052) Homepage
    Strange. The standard is to be "six miles" over single mode, and "328 feet" over multimode.

    I don't get it!

    I mean, we KNOW all decent standards use metric measurements - and Americans are inclined to convert them to the National Stupid System, so 328 feet makes sense (100 metres) - but where does this "6 miles" business come from? It is only 9,660 metres (9.66 km).

    Surely the standard will be 10,000 metres - ten kilometres, and the poster was lazy, and couldn't be bothered with the extra 0.2 of a mile?

    My question is this: when the specification is clearly based on very simple numbers: 100 metres and 10,000 metres - why convert that into the Stupid System? /.ers are not so stupid as to have to be fed figures fudged for obscurity!

    • Re:I'm going to guess... (Score:4, Insightful)

      by Ajehals (947354) <andyhalsall AT ictsc DOT com> on Tuesday December 05 2006, @12:49PM (#17115534) Homepage Journal
      Chicken and egg -

      Once the connectivity is there, hardware will become available and gradually more accessible as it is taken up, same goes the other way - if someone suddenly comes up with a bus and card capable of even higher speeds, it will slowly become available and more accessible until connectivity catches up and everyone wants it. Its all about getting to the point were a (potential) mass market appears and it makes the R&D viable. In the short term you will obviously see niche markets for it anyway - and they will pay buckets of cash for this kind of tech because they see a benefit from it.

    • Re:I'm going to guess... (Score:5, Funny)

      by jeffmeden (135043) on Tuesday December 05 2006, @01:37PM (#17116160) Homepage Journal
      Hi I'm Progress, and I'm going to guess that we haven't met. I will be forever pushing forward with faster speeds. Thought you were happy with Gigabit over copper? LUDDITE! 10Gbit is enough for all your communication needs since you can xfer the library of congress 5 times a minute? THINK AGAIN! 100G Ethernet is the natural progression and before long you WILL want it. Trust me, I have been working this way for thousands of years. Glad we could get acquainted, now excuse me I need to get back to hiding from politicians.

    • Re: (Score:3, Insightful)

      by forkazoo (138186)

      That off the shelf hardware won't be able to saturate a 100Gb connection.

      Depends on which shelf.

      Seriously, a lot of folks commenting on this news item seem to be convinved that all networks have only one node. Sorry, but I'm on a university, and I think that our interbuilding connections could really saturate a 10 Gb connection in the near future. It may be a long time before one PC can make use of a 100 Gb connection, but it won't be long at all before 1000 PC's can. Deployments will start the same way

All trademarks and copyrights on this page are owned by their respective owners. Comments are owned by the Poster. The Rest © 1997-2009 Geeknet, Inc.