Slashdot Mirror
IEEE Launches 400G Ethernet Standards Process
alphadogg writes "The IEEE this week launched a study group to explore development of a 400Gbps Ethernet standard to support booming demand for network bandwidth. Networks will need to support 58% compound annual growth rates in bandwidth on average, the IEEE claims, driven by simultaneous increases in users, access methodologies, access rates and services such as video on demand and social media. Networks would need to support capacity requirements of 1 terabit per second in 2015 and 10 terabit per second by 2020 if current trends continue, the organization says."
It says networks need to support it not individual machines.
Is it me or is the amount of information, when I look back through the history of the internet, that I get out of the 'net pretty much the same, just the traffic goes up?
Is all that bandwidth really just wasted on shiny?
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
When it comes to end users, gigabit is good for users that need to move large files, but 100baseT is plenty for the majority of desktop connections. I have many users that wouldn't notice if they were on a 10mb link because all they do is email and access a few lightweight browser-based apps.
However, on the server side of things, we struggle with only having gigabit. Unless you have a full mesh network, you need to think in terms of core enterprise infrastructure where the backbone could be handling transfers between dozens or hundreds (or thousands) of devices simultaneously. Once you get a few hundred servers talking, I bet you could saturate a 400gbps link pretty quickly. It's not intended for connecting to every single desktop PC out there.
I can't cite it, so it never happened, but the transfer of data, its more intensive examples, benefits corporations and governments and corporations and governments only. Human to human contact, such as voice calls, were promised to be ubiquitous and free because what worth corporations would derive from digital technology's rapid growth dwarfed what benefit an individual might. Instead, a text message is given charge by the character. An international call is distinguished from a local one. Maybe somone smarter than me can apply the addage about moving tapes in a station wagon to this latest mind blowing transfer rate and suss a conundrum of data capacity: Its storage versus its distribution. Whole library of congresses shall be traded between great power in the blink of an eye while the rest of us are to have thin clients driven by ad revenue.
Would were! Should is! Could be! And live a hundred times three.
Just looking at the current rate of growth and extending it out indefinitely is clearly absurd. Exponential growth has to run into physical, or financial, limits sooner or later. Sooner in this case, I think.
Anyway, where I am, there has been no increase in speed for the last 10 years. In fact it's gotten worse, as more and more people hook up and try to stream video on their phones, pads, etc., etc., but the upstream capacity is still the same and the phone company just shrugs knowing we have no choice. So if I applied the same naive style of prediction, I'd say we are going to have 1.8 MB/sec forever. (Which may be true, unless fibre is activated in my lifetime.)
It also remains to be seen whether the IEEE wants to go after some of the non-ethernet interconnects with this one, to try to get ethernet into use for larger-than-single-chassis interconnection of things that are usually confined to single boxes and 'internal' busses.
Your end user probably doesn't even need 1GbE; but his boring cheapo desktop probably has an 8(if 2.0) or 16(if 3.0) GB/s PCIe connector available for adding a graphics card. Hypertransport or QPI are faster still.
If one had the desires of people building larger-scale closely interconnected systems in mind, a very, very, very fast flavor of ethernet(with convenient ethernet features not generally available on internal busses, like the more sophisticated switching and routing capabilities); but enough speed to serve as an interconnect for a rack full of blade modules with virtualized storage and networking, or NUMA across all blades, or both, could be quite handy.
Such features have been available for a while in proprietary busses from the very expensive supercomputer outfits; but the IEEE may be looking to move in to that area with at least certain flavors of ethernet....
As of my post there were 8 posts, all pessimistic either stemming from "they will never be able to do it" or "customers wont want to upgrade" or "most of my customers are still 100mb, and thats all anyone will ever need"
Who are you people? This is a cool and exciting new technology. Since when did this become a website full of luddites? (and seriously, the "100mb/640k is enough for everybody" people can go fuck yourselves)
I'm a good cook. I'm a fantastic eater. - Steven Brust
FTFA:
The article also notes that 100G, which was ratified in 2010, is just now barely coming online.
Thus doing a little math, we're likely to see this standard in 2020 at the earliest, later if the nation collapses in insolvency.
Why are we shooting for 40 mpg vehicles when we know we need 100+ mpg? Hell, why doesn't Ford just go ahead and create a vehicle that can get 100,000 miles off a single tank of gas?
Which makes this perfect for datacenters...
And I am glad to see that some of the readers have picked up on that.
Not needed for desktop- as many have said, most have 100mb at home and wouldn't see a difference between that and GigE.
But infrastructure that was blazing fast 10 years ago with 100 users is now crawling at a snail's pace with 5000 users.
Usage per device has gone up quite a bit, which has an impact.
The increase in the number of connected devices has had an impact.
Add the two together....
Yeah, current network infrastructure is not sufficient for future needs and if this shit isn't worked on NOW we can all expect our broadband connections to perform like the dialup in the 90s a few years from now.
This is for enterprise and ISPs. Most of the equipment that uses this kind of bandwidth just splits it up and sends it on its way. Imagine the trunks that connect ATT to Sprint... They aren't going anything with the data but routing it. Check out this switch, and it's an old one: http://www.tech.proact.co.uk/foundry/foundry_bigiron_rx16_switch.htm
What feeds that? Trunks like we're talking about here.
400G is the next step up in the OTN hierarchy, it would be OTU5.
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0
There are more and more options for fiber in the USA now. You may need to move, but there are options and in places with typically good job markets.
If you're trying to compare 100GigE and above to single SSD drives, then you don't operate in the technical space these speeds are built for at this time.
Even corporate backbones bump into bottlenecks on occasion and I assure you that top end SANs can easily push that much data over a single interface considering they might have hundreds of drive in a massive array with caching technology that can bump performance even higher. And that's not considering if the drives are SSDs themselves.
And that's not even discussing servers that might have Fusion-io cards or something similar in them allowing huge I/O speeds out of a single server.
I would definitely have use for this, 1Gbps ain't nearly enough. 10Gbps would probably suffice, but those devices are horribly, horribly expensive and no one here in Finland seems to sell em for home-users at all.
And what makes you think the 400 gig devices are going to be any cheaper than the 10 gig?
You need to work on your reading-comprehension as I never said that. I only said that I'd have use for this. In fact I assume that 10Gbps devices will drop in price once 400Gbps hits the market.
There are very few 10 gig "home" devices anyway, it's probably going to be many years before there are any 400 gig "home" devices.
Doesn't negate anything I said.
Even SSD drives couldn't send data fast enough for this. Most of my customers still use 100baseT. Some have upgraded to gigabit. I see very little use for this outside of large data centers,
1) Computers in the future may weigh no more than 1.5 tons. - Popular Mechanics, 1949
2) I have traveled the length and breadth of this country and talked with the best people, and I can assure you that data processing is a fad that won’t last out the year. - Editor of Prentice Hall business books, 1957
3) There is no reason anyone would want a computer in their home. - Ken Olsen, 1977
4) We will never make a 32-bit operating system. - Bill Gates, 1989*
5) I believe OS/2 is destined to be the most important operating system, and possibly program, of all time. - Bill Gates, 1987
So... where do you think you rank? Source.
*Of course, you could argue that they didn't "make" it.
You seem to be forgetting about entertainment.
An HD video can be quite large. You start getting many people downloading or streaming them, and pretty soon you are going to need huge bandwidth.
Marketing buffoons: "Woohoo! More blinly-twirly CGI widgets and streamed kewt kitteh ads!"
Web Developers: "That will be $$$, please."
Data Center Professionals: "It's about time!"
Consumers on throttled DSL or cable connections: "Meh..."
Scruting the inscrutable for over 50 years.
This is what I'm talking about. These links will be buried deep inside infrastructure. Most of us will never see even one. How many of us have even seen a 10G link?
Moving uncompressed HD video (4:2:2 10-bit) requires about 1.5 Gbps, so I am very happy to see the ability to carry 266 professional video streams in one 400 GbE connection in the broadcast plant.
UHDTV1 (sometimes incorrectly called 4K) resolution at 60 fps requires 12 Gbps for 4:2:2 10-bit uncompressed, so it already jumps into 40 GbE connections. I have to admit I am not sure if we will see uncompressed 4K very often even in production, but potentially a visually lossless codec around 1 Gbps would make a lot of sense if it holds up through multiple encode/decode cycles.
Soon 100mbps will be dead in Australia at least.
With 50mbps and 100mbps internet plans readily available you'd need gigabit at least to avoid the internet slowing down your network or vice versa.
Why just 400Gbps if they figure they need 1Tbps by the year after next?
It's down to what is possible in the next few years. 100G was originally implemented as 4 lanes of 25Gbit/s, which was challenging on the electronics side. There is also now a cheaper technology with 10 lanes of 10Gbit/s. To get further you need both more parallelism and higher speed serialization-deserialization. However, increasing either of these numbers comes with a cost. 400G looks possible with 16 lanes of 25Gbit/s, but an increase to 25 x 40Gbit/s would be very difficult indeed. Here's a link to a NANOG presentation - http://www.nanog.org/meetings/nanog52/abstracts.php?pt=MTc2MSZuYW5vZzUy&nm=nanog52
I have. Many of you working in core IT will soon if you haven't already. They are all over the place in the heart of the biggest networks. This is because of the way common network architecture is done. Most networks at major corporations or institutions have a central core of some sort where all the VLANs run. That core is typically carrying traffic from most of the network segments all over the company. Sure, local traffic out at remote sites won't be going back there, but most of the server traffic in a headquarters datacenter will be running through a core like that. And when you have hundreds of servers hitting the core, nothing but 10G will cut it, which is why it is becoming so common in the heart of the corporate datacenter.
So it really comes down to what you do. On the client side you won't see this for a very long time, if ever, because most clients don't even use 100 MB in most circumstances. But it's all over for core corporate IT.
Beware of bugs in the above code; I have only proved it correct, not tried it.
I don't think anyone is currently suggesting this for a connection to a single node. It would be used as a switch fabric and at some point likely as a long haul link.
You do everything in RAM and on the GPU. Bitcoin mining here I come!
Defining Statistics and Social Research
Fax machines are still used in some fields because their acceptance as legally-binding copies of signed contracts has already been tested in the courts and case law / precedents already established. This has not yet occurred for "electronic signatures", so the legal validity of electronically signed contracts is not as well established in courts of law, at least in the USA.
.
There are also privacy issues, and the risk and susceptibility of interception when transmitting unencrypted sensitive information over the internet (like those pictures of checks). Wikipedia's article on "Fax" has this: Fax machines still retain some advantages, particularly in the transmission of sensitive material which, if sent over the Internet unencrypted, may be vulnerable to interception, without the need for telephone tapping. In some countries, because electronic signatures on contracts are not recognized by law while faxed contracts with copies of signatures are, fax machines enjoy continuing support in business.[2] In Japan, faxes are still used extensively for cultural reasons