Google's Subsea Fiber Optics(cloud.google.com) |
Google's Subsea Fiber Optics(cloud.google.com) |
Some years ago I bought a dozen extra copies to give away. Just last weekend I was hiking with a few friends. One mentioned that I'd given him a copy, and he started telling me stories from it. Especially what a revolution it was when you could communicate across the world in an instant.
We got to talking about the history of undersea cables. I mentioned a very cool interactive map of the current fiber cables and promised I would look it up and send him a link.
So, before I forget, here it is:
* https://www.youtube.com/watch?v=nKeZaNwPKPo
APNIC/NZNOG had a good presentation focusing on sub-sea optical stuff (January 2020)):
* https://blog.apnic.net/2020/02/12/at-the-bottom-of-the-sea-a...
For longer distances (>100km), you want to do a search for "coherent optics".
https://www.laserfocusworld.com/fiber-optics/article/1419605...
https://www.ofsoptics.com/wp-content/uploads/Hollow-Core-Fib...
In particular I always found the discussion about dealing with failure of global infrastructure that’s sitting in the ocean floor particularly intriguing. One non-obvious (to me at least) aspect was what do you do if there’s a failure of the power feed equipment that’s driving the cable. In most cases the answer was “double the voltage until you can repair it properly”. He explains it far better than I ever could on his blog: https://perspectives.mvdirona.com/2017/01/cs-responder-trans...
* An array of MEMS micro-mirrors which turn to reflect light, or
* Finisar's LCoS based method, which uses a 1" LCD TV display to display a gradient pattern which literally bends light.
It looks like the undersea cables use Finisar's (now II-IV's) WSS. I think the idea of bending light under software control is pretty cool.
Finisar's white paper showing optical path and other possible uses is here: https://www.amstechnologies.com/fileadmin/amsmedia/downloads...
Yet, the private sector focuses mostly on the software part, or services. I have rarely seen a start up on improving optical fiber or electronic chips. The public sector builds the infrastructure, often following decades of investment and work. People working on infrastructure either work for the government for pennies or, if they haven’t yet lost their jobs to outsourcing to developing countries, have difficulty finding employment. The profit goes to consumer companies focused on software or services; worse, these companies claim credit for the whole Internet.
Obviously CapEx will be large for a company with a product on infrastructure; there are monopolies; customers will be large operators, etc. Still, are there resources to better understand this issue? It always seemed to me a scam.
Also, will the situation change for “hardware”startups/companies?
I'm sure they only take in a fraction of that, and their costs are substantial. But even so... cloud bandwidth is overpriced.
There's lots of niche markets out there.
A letter in Morse code is made of up to four “dits” or “dahs”. Why would it take more than two minutes to send one letter?
https://www.google.com/books/edition/Report_of_the_Joint_Com...
I love scuba diving and I've personally seen many coral reefs wrecked from boat anchors and fishing nets, do undersea cables do similar damage? (I'm guessing yes).
Rocket launches typically develop and support bleeding edge technology. And are relatively infrequent compared to regular travel flights.
The "huge" plastic cables are 7-8cm (~3 inches) diameter. So big, but not gas pipeline big. And because of THEIR existence many people are NOT taking those regular travel flights they otherwise might.
Not saying we can't do better. But people closer to it and smarter than me would know.
This looks forced and cringey.
Yes! That everything is fine, everyone is happy and if you're not happy, then the only sane conclusion is that there's something wrong with you.
The ends of the fiber might be at different depths with a pressure difference that could move water a long way into the fiber. I imagine the length that water got into would be ruined even if the water was pushed out again.
I conjecture that undersea hollow-core might end up being expensive to maintain.
I imagine that you need less of them, because "faster" also means longer range until attenuation kicks in. How much more range/less amplifiers compared to current state of the art I don't know. I'd think it should be possible to lay down a new part the whole length between two amplifiers/laser pumps/couplings/ and "plug" it in by remote controlled submersibles, instead of lifting up the broken ends, and "splicing" it on the ship. Like pre-made Cat-5/7 with RJ-45 plugs, so to speak :-)
And, a hole that small in a block of glass could withstand a titanic amount of pressure.
The more important figure is the attenuation at frequency bands used in telecommunication (C-band and L-band). According to the article, hollow core fibers are still much worse than solid core fibers (0.28 dB/km vs 0.14 dB/km).
No, pretty sure light still travels through them at C. What they can do is carry more data, largely by having lower error rates.
It does? I don't think I follow. In your list of components, every single one of those is, at least in the US, largely or almost entirely handled by private companies.
The big semiconductor companies are private. I actually don't think there are any notable public entities that make their own chips. Hardware companies (I'm assuming you're talking about things like motherboards, routers, switches, etc) are private. Fiber optics/communications/networks are laid almost entirely by private telecom companies (and there's actually a big push to take this away from private companies and make ISPs be government entities). The article that you're commenting on is all about a private entity investing money into laying fiber and improving the protocols that communicate over it.
>I have rarely seen a start up on improving optical fiber or electronic chips.
There are a lot of SMBs working on chip design. I'm less familiar with fiber, but a quick google shows at least a couple, all private.
Consider high speed optical communication. The invention of transistor, laser, optical fiber, optical amplifiers, modulators, communication and information theory, DSP, and many other important technologies, was funded by taxpayers for a long time. You may say, some of it happened in old bell labs, but bell labs was a state-subsidized monopoly, rather an exception, and even things like information theory were really developed at MIT and other public research labs (Shannon had an office at Bell Labs for some time but had limited contact there).
Consider machine learning. From 1940s to 2010, neural networks were funded by governments grants, and developed at universities. The governments even funded building practical applications out of deep neural networks, for example, in the area of speech recognition, unsuccessfully. The periods are now called AI winters. It’s really in the past 15 years that the subject has become practical and companies are building products based on it. FANG is capturing the profits of this decades-long research and development.
Today there are millions of graduate students and postdocs and professors at universities performing risky research and experiments. A lot of it doesn’t pan out. The public bears the risk. Once an idea begins to work, it’s handed over to private sector for further development and bringing the work to market.
Most of the research papers published at conferences and journals, Nobel prizes, Fields Medalists etc come out of universities, not startups or FANG.
Yes, Google installed fiber cables (perhaps even bought it from Corning or the like, and also contracted out the installment). But that’s not infrastructure R&D.
I think there is no denying that basic long term research is done by and large in public sector. Private sector is still focused on short term product research.
Citation desperately needed. How you got "the government pays for our internet infrastructure" out of an article about Google paying for a new Subsea cable, I do not know.
I implore you to do a tracert right now to hackernews, and lookup who operates the ips of every router it hits in between. Chances are incredibly good not a single hop is on a government network. (true for the U.S at least)
Having said that, what you've said is basically wrong.
> I have rarely seen a start up on improving optical fiber or electronic chips.
Unless you are trying to be pedantic by saying it has to be a startup rather than a private company this is completely wrong.
TSMC, Intel, Samsung and IBM do most of the work improving electronic chips manufacturing processes.
I'm not very familiar with optical fiber work, but I know NTT in Japan does a lot of R&D in the area.
How much of VC funding goes to companies focused on services or software vs hardware or infrastructure?They actually write down their areas of focus on their websites. I don’t see many of them stating they fund companies developing better chips, optical amplifiers, fibers or other pieces of infrastructure. It’s mostly services, software, apps, consumer applications, and similar.
And to my point, how much do you make if you are a hardware engineer at Intel (a large hardware company), vs a software engineer at say Google (a large software company) with similar years of experience?
It's a little off your point but there was an amusing book published a few years ago which went through twelve, from memory, technologies on which the iPhone was totally dependent and which had all been developed out of the public purse.
Can't remember what it was called so if anyone recognises it I'd be interested to hear.
I haven't read it, but from the TOC, it looks like these are the technologies that are "invented" by public sector that the iPhone (and other smartphones) rely on:
* Giant magnetoresistance (GMR), SPINTRONICS programme and hard disk drives
* Solid-state chemistry and silicon-based semiconductor devices
* From capacitive sensing to click-wheels
* From click-wheels to multi-touch screens
* Internet and HTTP/HTML
* GPS and SIRI
* Battery, display and other technologies
ref: https://cms.marianamazzucato.com/wp-content/uploads/2021/04/...
Summary HBR article: https://hbr.org/2013/03/taxpayers-helped-apple-but-app
Having said that, cloud bandwidth is indeed overpriced; but at the same time, given that Google Cloud is still burning money, can it perhaps be argued that bandwidth is one of the money makers that allow for other services to be free?
I recall that from the old webhosting days, this was already a common tactic of the providers: lure people in with cheap servers, sometimes even at a loss, and earn money back with bandwidth.
Resources are used depending on prices. If the costs for providing bandwidth are low and everything else is expensive, but the prices are the other way round, then the economy optimizes to waste resources. That's not sustainable.
Furthermore, software (as a product/service) has the lowest marginal cost of nearly any product. Given the cost it takes to have one more customer on your platform is some nominally small amount of bandwidth (which depending on the product, can be sub 1 gigabyte per month) the additional expense is easily justified.
"""These features allow many B4 links to run at near 100% utilization and all links to average 70% utilization over long time periods, corresponding to 2-3x efficiency improvements relative to standard practice"""
† The video said this started five years ago, so there appears to be a lot of lead time that is needed.
They do invest in new chips. There's a bunch of "AI" chips funded by VC which are focused on the outside returns available by improving training performance. These are all funded by private investment.
In addition there are feedback, failover and monitoring functions that require more optical components to be included, and it's likely that this type of functionality will increase as demand for improved latency and reliability increases, and new cable networks are built.
Edit: cf.
[0] https://arstechnica.com/information-technology/2016/11/priva... [1] https://www.reuters.com/article/us-highfrequency-microwave/l...
They will have servers “next door” to the exchanges, but need the servers to have incredible low latency connections to each other.