If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Ars Technica)   Meh: New fiber optic cables. Fark: 800Gbps fiber optic cables   (arstechnica.com) divider line 17
    More: Spiffy, Intel, throughput, supercomputers, data center, Fujitsu, certification program, optical systems, CPUs  
•       •       •

1521 clicks; posted to Geek » on 11 Mar 2014 at 6:50 AM (19 weeks ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



17 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest
 
2014-03-10 11:51:34 PM
Wow. That's almost as fast as Korea's home internet speeds. Maybe America will finally have a chance to dominate Starcraft tournaments.

kekekekekekeke
 
2014-03-11 12:40:51 AM
Just to have Comcast throttle you to 6 Mbps.
 
2014-03-11 07:29:02 AM
I'm meeting with the techs from my city's fiberoptic network today.  I plan on showing this to them, but I know they're already eyeing 10-Gbs on a ~3-5 year window.

/you can buy gigabit from my city right now for $150/month
//the worst possible internet connection they sell is 20/20 for $35/month
 
2014-03-11 07:39:58 AM

ecmoRandomNumbers: Just to have Comcast throttle you to 6 Mbps.


At least you can still watch Netflix in HD.

/fios
 
2014-03-11 07:40:06 AM
This seems cool in a "maybe in 5 to 10 years" sort of way, but I just can't imagine widespread deployment of this tech any time soon. First, you can add it to existing lab infrastructure, but that kind of bandwidth seems wasted outside of backbone applications. Even then, 40GbE is seeing very slow adoption due to cabling complexity and per-port cost; if data center managers (and CIOs) are still reluctant to eat those costs for 40Gb bandwidth, I'd give this more than a decade before it hits mainstream. It might be useful for super-high-bandwidth conections between nodes in a HPC cluster, but InfiniBand already has a solid hold there and I'm not sure what the difference is in latency.

Okay, now I'm bored and you probably are too.
/rant over
 
2014-03-11 07:54:28 AM
That's cool but I'll wait for the Monster ultra diamond quadruple shielded version.
 
2014-03-11 08:11:25 AM

That Guy What Stole the Bacon: This seems cool in a "maybe in 5 to 10 years" sort of way, but I just can't imagine widespread deployment of this tech any time soon. First, you can add it to existing lab infrastructure, but that kind of bandwidth seems wasted outside of backbone applications. Even then, 40GbE is seeing very slow adoption due to cabling complexity and per-port cost; if data center managers (and CIOs) are still reluctant to eat those costs for 40Gb bandwidth, I'd give this more than a decade before it hits mainstream. It might be useful for super-high-bandwidth conections between nodes in a HPC cluster, but InfiniBand already has a solid hold there and I'm not sure what the difference is in latency.

Okay, now I'm bored and you probably are too.
/rant over


That's what this is designed for.  This isn't for long distance stuff, this is for local interconnects inside of clusters and data centers.
 
2014-03-11 08:12:11 AM
Eh, there's going to be a huge bottle neck with your PC's IO, no matter how serious your rig is. It'll be a while before the PC can take full advantage of fiber optic.
 
2014-03-11 08:32:43 AM

UNC_Samurai: I'm meeting with the techs from my city's fiberoptic network today.  I plan on showing this to them, but I know they're already eyeing 10-Gbs on a ~3-5 year window.

/you can buy gigabit from my city right now for $150/month
//the worst possible internet connection they sell is 20/20 for $35/month


Wiiiiiiiiiilllllllson?
 
2014-03-11 08:33:42 AM

TuteTibiImperes: That Guy What Stole the Bacon: This seems cool in a "maybe in 5 to 10 years" sort of way, but I just can't imagine widespread deployment of this tech any time soon. First, you can add it to existing lab infrastructure, but that kind of bandwidth seems wasted outside of backbone applications. Even then, 40GbE is seeing very slow adoption due to cabling complexity and per-port cost; if data center managers (and CIOs) are still reluctant to eat those costs for 40Gb bandwidth, I'd give this more than a decade before it hits mainstream. It might be useful for super-high-bandwidth conections between nodes in a HPC cluster, but InfiniBand already has a solid hold there and I'm not sure what the difference is in latency.

Okay, now I'm bored and you probably are too.
/rant over

That's what this is designed for.  This isn't for long distance stuff, this is for local interconnects inside of clusters and data centers.


Understood, but the second part of that sentence was about IB and latency. End-to-end IB latencies tend to be lower than optical latencies (yes, by a tiny fraction, but it can make a difference), so I suppose I'm just playing devil's advocate when I consider that it may be more cost-effective to use existing IB infrastructure in a higher density than to switch to an entirely new infrastructure. Granted, I'm sure different applications and needs come into play depending on a given setup, but I like to be skeptical of new tech. I'm just a jerk that way, I guess...
 
2014-03-11 08:39:30 AM

Shakin_Haitian: UNC_Samurai: I'm meeting with the techs from my city's fiberoptic network today.  I plan on showing this to them, but I know they're already eyeing 10-Gbs on a ~3-5 year window.

/you can buy gigabit from my city right now for $150/month
//the worst possible internet connection they sell is 20/20 for $35/month

Wiiiiiiiiiilllllllson?


Yep, we're known for three things - the best damn BBQ in the world, at one time the largest tobacco auction market in the world, and now our internet is so good Slime-Warner crapped their pants in fear and ran running to the state government to make it illegal.
 
2014-03-11 08:40:23 AM
So which state will be the first to outlaw this new fiber optic technology? My money's on Kansas.
 
2014-03-11 09:22:11 AM

That Guy What Stole the Bacon: data center managers (and CIOs) are still reluctant to eat those costs for 40Gb bandwidth


Which is why we're still stuck in the last century. It's going to be decades, as you predict before people are born into data stream culture and begin to demand a better environment. It's life quality. Right now, it's being run by idiots who are used to waiting for shiat. People who would pick up a telephone and wait for a dial tone, then begin to conduct the business of dialing. Once they all die off, me included, we might see improvements.

Too many dinosaurs right now.
 
2014-03-11 10:59:59 AM

vudukungfu: That Guy What Stole the Bacon: data center managers (and CIOs) are still reluctant to eat those costs for 40Gb bandwidth

Which is why we're still stuck in the last century. It's going to be decades, as you predict before people are born into data stream culture and begin to demand a better environment. It's life quality. Right now, it's being run by idiots who are used to waiting for shiat. People who would pick up a telephone and wait for a dial tone, then begin to conduct the business of dialing. Once they all die off, me included, we might see improvements.

Too many dinosaurs right now.


Agreed, to an extent. Another aspect to consider is that those of us lamenting the state of American connectivity aren't really the majority of consumers. ISPs over here have done a very good job of training the market to accept, as "bleeding edge," technology that's years old over in Europe and Asia.

Sure, Hong Kong's average per-connection bandwidth was around 60Mbps last year (at least according to Bloomberg), but why would an American ISP offer that to a market impressed with an 8.5Mbps average that they already pay an arm and a leg for?

To be fair to the dirtbags, it makes excellent business sense: get your market used to paying a premium for a product that's obsolete and has been put through its paces in other markets and thus cheaper for you to implement and provide on your own schedule. Maintain that trend, and what cost hundreds of millions to implement in another market when it was new may only cost you tend of millions to implement when your domestic market is "ready" for it. And, since most of your market has no idea how far behind they really are, you can still sell the service to them as cutting edge, priced accordingly.
 
2014-03-11 12:45:46 PM
Cool: this

Not cool: vendor caps

supply & demand
 
2014-03-11 03:38:06 PM
Wake me when we have isolinear chips.
 
2014-03-11 09:46:16 PM

ecmoRandomNumbers: Just to have Comcast throttle you to 6 Mbps.


Wow, I've got Windstream and I'm up to .4% of 100 Mbps.  If I'm lucky I'll hit speeds of .75% of 100.  It's like having a dial-up without the whining.  God I miss Taiwan.
 
Displayed 17 of 17 comments

View Voting Results: Smartest and Funniest


This thread is closed to new comments.

Continue Farking
Submit a Link »






Report