Slashdot Mirror
Why We Don't Need Gigabit Networks (Yet)
AmyVernon writes "Most computers today can't support gigabit connections and current Wi-Fi networks can't offer those speeds either. The first trial of Sonic.Net's gigabit network was a speed test on a generic laptop that showed off 420 Mbps down; the laptop couldn't handle a full gig. Plus, few applications need those speeds. It's hard to justify such a huge investment in a network that will have few subscribers and few applications that need it. Of course, that can change, and then these networks will be vital. This story has a good analysis of where things stand and what has to change."
Gigabit networks are important when working with almost any kind of file copy. I am not sure the last time someone tried to backup even just 100GB of data (Think backups) over a 100 megabit network. Copies like that can take for ever a fully saturate 100 megabit network and slow down traffic for everyone. While copies over gigabit rarely use the entire pipe its good to know that there is still bandwidth left over for other tasks.
How DARE you say we don't need faster networks! This article should be purged from the interwebs and timothy should be strung up by his gonads for even considering posting it!
I have 2 computers, a ps3, and a wii connected to the net. even if I am doing something simple like streaming a movie from one computer to the ps3 to watch on my tv while someone else is playing a game online, downloading something or the other, or just generally using the web to watch anything in HD, I could easily find a use for that bandwidth.
My sentiments exactly. If you build a (vastly) faster internet, the hardware will be ready for it long before you actually have it to the customers doors. And yes. we do need it. Why? because the vast shift towards streaming all media content is going to murder the shit out of the current system. If a customer can download an entire HD movie in seconds, that frees up the network in general, instead of slowly moving the file over the course of an hour. The streaming model can die the death it deserves at that point.
I've decided to Diversify my Holdings. I've divided my cash between my left and right pockets, instead of all in one.
The reason this post is stupid is that infrastructure is long term. When you go to the trouble of sending out a crew to dig up and put fiber in the ground your putting in an infrastructure asset that should have a 15-30 year lifespan. The fact that can average machine can't saturate it today means we're being forward thinking.
... and 640K computers.
Build a faster network and someone will invent more devices to connect to the network to shove around data that they don't need.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
not everyone keeps/needs a server at home.
The misconception that people would have no use for a home server has led to prevalence of highly asymmetric Internet connections suited more to a spectator culture than to a participatory culture. Have you heard of the FreedomBox?
FTA:
"So we’re stuck at a point where a gigabit — or even 100 Mbps – sounds awesome, but it’s not exactly worth the prices most companies want (or need to charge). This is why Google’s and Sonic.Net’s plans to expand moderately priced 100 Mbps and gigabit networks will be so important."
The summary to this article is misleading. It led me to write a mini-rant about the usefulness of gigabit LANs. In fact, the article's talking about gigabit WAN connections at the home. Their denouncement has the tinge of that old Microsoft exec quote about the internet being a fad and no one needing very much ram.
FTA:
“If every consumer has 100 Mbps, we’d have some better applications,” Jasper said. ” At 100 Mbps, high-def video conferencing becomes a reality and you don’t need local storage anymore. You don’t even need local computing.”
You went from talking about gigabit WANs (at the corporate level), to the use of fast ethernet WAN at home. Somehow, there's a use-case at the home that isn't there at the corporation.
And this made /. frontpage, why? Can I get a +5 comment simply by using the words "100 mbps, gigabit, ethernet, 802.11[n-z], important, high-def, local, storage, computing" ?
PS: I don't reply to ACs.
is not for one computer to saturate it, but for 10 machines to get decent throughput simultaneously.
As the island of our knowledge grows, so does the shore of our ignorance.
When you wake up to the obvious facts of 1999, let me know, and I'll give you an invite to the 21st century. Cuz I'm k3vvL and rollz like dat.
Shoes for Industry. Shoes for the Dead.
The streaming model can die the death it deserves at that point.
Streaming is still the only model I can see for live events such as news talk shows, sports, scripted sports (e.g. WWE PPV), concerts, and the like where viewing begins before the whole video has even been recorded.
There are about 200 people who use my department's network at any given time during the day, and maybe 50 at night. All the desktops have their filesystems mounted on NFS, and people routinely upload or download large datasets. Gigabit networking is not even fast enough for what we do (yet somehow we have trouble getting that much installed).
Palm trees and 8
For the software end of this, check out CrashPlan. It saves incremental backups of your system to external hard drives, your friend's computer (also running CrashPlan) and/or the CrashPlan servers. It's great stuff, and works on Win/Mac/Linux. Plus, your backup data is encrypted before it leaves your computer, so you don't have to worry about the security of your friend's computer. (By default, your data can be decrypted on the CrashPlan server in order to support web access to your files. If you don't want that, you can set an encryption password that CrashPlan can't access, and then no one can see your data outside of your computer.)
Eh.. nothing wrong with the streaming model, as far as network bandwidth goes, and there are advantages for things that lots of people are trying to get: under the streaming model, you can use multicast a little better. For instance, if something just needs to be downloaded, you can multicast it in a continuous loop and clients can assemble the pieces in the right order at their end, waiting until they have enough to finish.
Or you can have multiple streams staggered (for say, video), so that people can join in at almost any time and get on a stream.
The other advantage of streaming is that if, say, a client is downloading video and the user decides halfway through to stop watching, you don't have to send any more bits.
Can you be Even More Awesome?!
A 100Mbps connection requires one 10GbE aggregation router for every 1,000 users, a 1Gig connection requires on for every hundred.
This is not so. As I wrote we have 1000 users on a gigabit with only a peak of 300 Mbps used. This is fact, that is the way our network operates. Everyone here has gigabit in their home.
So yes I have little hands on experience with the "industry" but I DO have hands on experience with a network with 4000 people and 1000 subscribers. Do you?
And yes the routers are expensive, but not THAT expensive. A router capable of routing 10 Gbit/s is NOT hundreds of thousands of dollars. We have a HP 5412 switch for our core and it has a backbone capable of 40 Gbit/s. As far as I remember we paid around 20k USD for it. Our edge switches are HP 2910 (60 of them) every one capable of 10 Gbit/s as well (two 10 Gbit/s ports and 48 1 Gbit/s ports). Those are 5k USD each (without the 10 Gbit/s option as we have in fact no use for that - 40 users are simply not generating that much traffic).
You are confusing edge routes with the real core routers of the internet. Those are already running 10 Gbit/s or more and would not need upgrading. Giving people access to higher speeds do not generally make them generate more traffic aggregated.
I think the point you are missing is this: Upgrading the edge DOES NOT MEAN THE CORE NEEDS UPGRADING.
No, but the argument is "this cheap and shitty laptop could only manage to use half of the gig connection, so therefore no one needs gig speeds for the home".
An argument that is easily destroyed by saying "ok, do you live alone? Do you have more than one person using a computer at the same time?"
It's not just servers. I share a house with 4 other people and we can all watch HD streaming video on the connection we have, just. If the bandwidth goes up a little, or people start using off-site backup more frequently I can see a market for a consumer-level gig connection. I know you can already get them in some other European countries (here in the UK, the best you can get on a consumer budget is 100Mb (soon to be 200Mb) from Virgin cable).
One shitty laptop might choke on a gig connection, but three or four computers will happily share it.
My organization is on the verge of needing to move our equipment to 10gig soon, because the 1gig network is starting to become a serious bottleneck...
Gee... what third world country or year is the OP posting from??? My workstation, built in 2007, supports GigE and is capable of speeds nearing the theoretical limit, just as it can on a 100Mbps link. I also have several of my servers connected on the same GigE VLAN. As for 10Gig... my old employer had racks and racks of servers which we tested and found to be able to use a significant portion of a 10Gig link. Of course, these systems were using NICs which were at the time (about 2 years ago) running about $700 ea., and connected to switches running about $20K each... but when you build a super computing cluster (to which we had a dedicated 10Gbps connection to the NLR)... you don't skimp on your infrastructure.
Helping build UN*X and the Internet since 1981.
Exactly. On top of that, there are a large number of perfectly sensible reasons why ISPs should be installing gigabit networks today:
Even if 1Gpbs is not materially more useful today than e.g. 100Mbps, we don't actually have 100Mbps connections. And any 100+ connection is materially more useful today than the existing 10-20Mbps connections. So if you already have to roll a truck to do an upgrade you might as well not half ass it, because you can save money in the long run by making it all that much longer before you have to do it again.
On top of that, even if nobody actually needs a 1Gbps connection right now, if you can offer one and your competitors can only offer 50Mbps then you capture a big marketing advantage. "20 times faster than Verizon FIOS" etc. It also puts you in a stronger position for the next black swan event. If somebody invents a cheap holodeck and it turns out to require 150Mbps per person to interact with others over the internet, you can cash in immediately while your competitor has to run around frantically upgrading their equipment while they hemorrhage customers to you.
I have heard this argument so many times before, but it is just stupid to say: You can only use 400MBit, so better keep your 100MBit instead of getting that full GBit, as you would not be able to use it fully. It might not be 10 times faster, but it still is 4 times faster, which might well be worth the price to some.
However, this being computer technology, "shit you don't need" might not actually cost anything, or it may be cheaper for you to just use the better tech.
For instance, with regular ethernet, the standard allows for 10, 100, and 1000 Mbit/sec. However, no one bothers making the 10Mb stuff any more, not only because it's slow, but because it's obsolete: it doesn't save any money to make a 10Mbit ethernet transceiver than a 100Mbit one, as the cost is in the silicon, not the speed. Even if 75% of customers didn't care, it wouldn't save the hardware makers any money to go back to 10Mbit hardware.
I can't speak for the telecoms/cablecos, but perhaps the situation is much like that for them too. From their PoV, all they really care about is 1) keeping customers happy so they don't switch to the competition (although there's usually only one competitor), and 2) the aggregate data demand, so they can size their backbone links accordingly. The speed of individual links probably isn't quite that important, except maybe for advertising, and there may not be any real cost difference in deploying one speed versus another. As we've seen in other computing and electronic technology, once the technology is out there, there's frequently no cost savings in sticking with the old version; in fact, the new stuff is usually cheaper than the old stuff, and the only cost is upgrading, which may be mitigated by power savings or other factors. So, for instance, if a cableco is building infrastructure to a new neighborhood, they can either install the latest stuff (DOCSIS3), or something older (like DOCSIS1). But is there really a cost savings in putting the older stuff in? They probably don't even make the older stuff any more, so they just install the latest. And on top of that, if they did install old, slow DOCSIS1 stuff, they'd probably have more problems with minor incompatibilities between that equipment and the customers' brand-new DOCSIS3 modems they just bought from BestBuy or NewEgg, which are tested by the mfgrs to work properly with the newer equipment, and may not have been tested thoroughly on old stuff that they can't even find any more.
Several years ago I was working on a networking product that had a couple of 1Gbps ethernet ports. We needed to test it at all the different speeds that Ethernet supports, to make sure it conformed to the standard, but we had a hell of a time trying to find a 10Mbps switch or hub. I think we managed to dig one up from an old storeroom somewhere, that we luckily hadn't thrown away yet. We could have manually set a port on our nice managed switches to 10Mbps of course, but that's really not the same as having hardware that really works that way natively.
Um, probably not. He's rolling out fiber in my town of 7000. But it's also his business to know whether it's really going to be saturated so they can do the right network on the backend. Sonic.net is a pretty kick-ass ISP. They instituted outbound SMTP blocking. But they noticed I'm running my own SMTP server and sent me an email saying they weren't blocking SMTP to/from me, but I could enable/disable it just by visiting my member account page. Also, they just rolled out free fax numbers (gateway to PDF/email) and outbound faxing for everyone.
Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
Of course the summary leaves out the part of the conversation where bandwidth is also a measure of well, bandwidth. Just because one single individual device can only get ~500MB does not mean that GbE is worthless. What if there are two laptops sharing that connection? It will be tapped out. Put another one on there and all of a sudden a one gig pipe is not big enough.
How stupid are people, really?
I read the article as a laptop being too slow because of its drive not handling data that quickly.
That neatly bypasses a very real need for high speed low latency remote connections where disk speed is irrelevant -- remote desktops, remote apps and VPN, often in combinations. And in combination with other things that suck bandwidth too.
There's more to bandwidth than file transfers.
it's also his business to know whether it's really going to be saturated so they can do the right network on the backend.
Naw, that's Jasper's Engineers' business to know. As CEO, it's Jasper's business to swim around in a pile of money like Scrooge McDuck.
Holy shit... as a developer of broadcast television infrastructure equipment which is used throughout the world, though predominantly in Europe, I thought I was clued in quite a bit on the back-assward methods used within the industry for transmitting TV and Internet signals.
This was a total shocker for me though. I'm a huge fan of using wavelength multiplexing within fiber. Especially when the fiber in my house is a single fiber as opposed to pairs which are much harder to make look pretty in a house. However this is one of the funniest things I ever heard of.
It's taken a really long time for the industry to finally come up with a less than insanely shitty method of using coax cabling for digital media access. Oddly enough, the cable companies have more or less completely rebuild their coax backbones to make it happen... what makes it odd is that they wanted to keep the coax to avoid having to lay new cable.. haha wow that worked well.
Now, it appears that Verizon has decided to transmit the entire cable multiplex over a single wavelength, therefore allowing them to a) guarantee their bandwidth usage even if it's insanely high, b) decrease hackability of TV fiber as it is on a not so common wavelength and therefore difficult for consumers/hackers to get receivers for it. c) run less expensive multiplexers they wouldn't require conditional IP multicasting at the switches. d) decrease the cost of maintaining a huge TCP/IP network of devices as it would be possible to remove the IP layer altogether and use a more reliable ATM style layer.
This design so fantastically screws consumers into buying/leasing equipment exclusively from Verizon that it damn near guarantees Verizon a minimum of $30 a month extra per average household just in equipment rentals. And what's best is, they can claim "Sure, we support using third party hardware with our system Mr. FCC, but there's no law that says we have to help anyone make equipment that works with out network is there? But if anyone ever does... sure, we'll support them".
The only true benefit of this design to the consumer is that it would be possible to make a fiber to DVB-C converter that would theoretically make it possible for a TV to receive the signal using the digital coax connection within the TV.
I am SOOOOO glad I don't live in the states anymore... this stuff would infuriate me... it's bad enough I had to make an FPGA for brute forcing DVB-CAS in order to cut my power usage in the house by 100watts (24/7 since the shutting off the set top box from the fiber company requires a 3-5 minute startup time). Now I use an FPGA which consumes 5watts to crack the keys and shared them out with the rest of the house. Saves me a fortune. The FIOS thing would drive me nuts... oh there's the additional bitch about FIOS which is that it's DOG SLOW!!!!