Google Buys a Piece of a Cable To Japan

Googling Yourself writes "Google announced that they will be part of a six-company consortium that will build a high-bandwidth sub-sea fiber optic cable linking the US and Japan. The new cable system, named Unity, is expected initially to increase Trans-Pacific lit cable capacity by about 20 percent, with the potential to add up to 7.68 Terabits per second of bandwidth across the Pacific. The name Unity was chosen to signify a new type of consortium, born out of potentially competing systems, to emerge as a system within a system, offering ownership and management of individual fiber pairs. Rumors that Google would join the consortium had originally surfaced in September last year but the company had declined to confirm or deny the news."

Re:Outsource is now faster!

[Idefix97]

If you RTA you see that an Indian company is involved in building the cable as well - it's NOT just Japan.
From the article: "Bharti Airtel Limited, is India's leading integrated telecom services provider with an aggregate of 60 million customers."

11.43 seconds per LoC

[Mage Powers]

I don't really understand LoC, so I googled a figure.
http://www.uplink.freeuk.com/data.html [freeuk.com]

10 Terabytes: The printed collection of the US Library of Congress

mage@prometheus:~$ calc 7/8
                0.875
mage@prometheus:~$ calc 10/.875
                ~11.42857142857142857143

11.43 seconds per LoC

Re:Obligatory

[Umuri]

Being as the standard estimate used in the LoC/s standard (20 terabytes/sec), and it's a 7.68 terabit line, you can do some simple math.

Assuming that a terabyte is 8 terabits, the line adds .96 terabytes/sec in bandwidth, so you'd get the Library of Congress in around 21 seconds.

So approximately 1/21 LoC/sec, or 2.85 LoC/min.

Nothing nefarious, just good business

[Raindeer]

Just posted this on my blog: http://lunaticthought.blogspot.com/ [blogspot.com] (You can find more info here on the economics of submarine fibre and nice pictures of the Tyco Responder Cable laying vessel)

Gigaom is reporting on Google buying a share into the Unity submarine cable. Many people will read into this an attempt by Google to become a telco or do anything out of its current layer 7 service and application business. I don't belief it is, it's just simple economics. Google now buys wholesale capacity instead of retail. My reaction on Gigaom was:

One of the main drivers for wanting your own fibre on certain submarine routes is the pricing strategy of the owners of the submarine fiber. Traditionally these fibres have been owned by incumbent national monopolists. Their pricing was set at a fixed price per Mbit/s. If your banndwidth utilisation grew, their income grew too, though their costs didn't, leading to excess profits. On the Transatlantic route this problem has been solved by having an oversupply of commercial competitive fiber. The oversupply resulted in a situation I call mutually assured destruction, where everybody went bankrupt and whole networks were sold for pennies.

On the Pacific route it's mostly incumbent national monopolists owning fibre and they probably have learned from the Atlantic disaster. This means prices don't drop (or not as quickly as traffic growth) and that means that some parties see an increase in their traffic costs. Google now has solved this by joining a club of submarine fiber owners and not having to worry anymore about the cost of a megabit/s. Google just has to worry about when they will fill up their terabit chunk and when someone will slice through the fibre.

BTW I'm willing to bet Google will join another club on this route to add some much needed redundancy.

Re:how many strands

[The Master Control P]

AFAIK, the highest speeds on a single fiber are 10Gbps, which would make this a bundle of 768 fibers (which would make sense in accordance with things involving computers commonly involving powers of 2 & 3).

IANAIE (internet engineer) though.

Re:how many strands

[OeLeWaPpErKe]

Actually you might want to look up DWDM and CWDM. It really depends on the cable, but even with systems commercially available (if a bit pricey) today, you can do up to 80 channels on a single fiber strand, allowing for 800 Gbit per fiber line (half duplex), which means you can achieve 800 Gbit per fiber pair full duplex.

This obviously depends on a lot of things, the most important parts are the retransmission stations on the cable (every 50 km or so), as they're very hard, and VERY expensive to replace. Generally half the fiber strands are backups, and all cables connect only to either even or uneven retransmission stations, allowing the cable to keep functioning with the loss of any one retransmission station (a frequent occurance). Problem is that for repairs a ship needs to come by, retrieve 3 transmission stations from their 200 meter depths, and get engineers close enough to conduct repairs, while preventing other ships from crossing the exposed 200-or-so kilometer of exposed cable. This is one of those ships [k-kcs.jp].

Btw these retransmission stations are sinking pods that "float" below the ocean at a given depth (generally 200 meters or so). They are powered by a high voltage current transmission system in the cable itself.

Wikipedia entry on WDM [wikipedia.org]

Re:Good news all around if it happens

[totally bogus dude]

I don't think it will have much, if any, effect on Australia. Most of the focus of our ISPs is getting to the West coast of North America, and going via Japan is a pretty significant detour. Mind you, the Australia-Japan Cable gives us 320 Gbit/sec to Japan.

A trace from a server in Ohio to 72.14.235.104 (one of google.co.jp's addresses) has a RTT of 200 ms, which is about the same as the East coast of Australia to www.google.com (I get 230 ms from Perth). So for US-based sites going via this new cable would, I imagine, be quite a bit slower than via more direct links. We also already have several independent links to the US, so it wouldn't even be much benefit to us as a backup.

According to this random site [happyzebra.com], Sydney to San Jose is almost 12,000 km (by air). Sydney to Tokyo is 7,700 km. The press release declares that the new cable will be approximately 10,000 km, so that's around an extra 5,000 km via this route minimum. I suspect any run from Australia to Japan isn't going to be particular direct though; AJC is apparently 12,700 km.

This PDF [atug.com.au] provides some maps of the approximate cable locations. It has one marked "New Japan-US Plans" which might be referring to the Unity cable.

Re:Well...

[tkh]

You should use more reliable information source on this. Google *is* the most popular search engine in Japan, and YouTube traffic has been taking bandwidth between the U.S. and Japan. This is one of the main motivations for Google to buy a piece of a cable.

Re:Bandwidth

[SeaFox]

More bandwidth for adverts?

No, more bandwidth for undersea-cable rape hentai.

Re:Good news all around if it happens

[daBass]

It will have a big impact, see my other comment [slashdot.org].

In short: we are getting a 2Tb cable to Guam in 2009 and Unity's Southern loop will go through there too.

Sweet!

[Sabz5150]

A brand new fat pipe to download tentacle he^H^H^H^H^H^H^H^H internet content!

Re:Well...

[fullback]

Yahoo is the most popular search engine in Japan.

Japanese speak Japanese. They don't visit or read English-langiage websites any more than Americans visit Japanese-language websites hosted in Japan. Get it?

I live in Japan. You know nothing.

Re:retransmission stations - power supply?

[jacquesm]

they use laser to pump the optical amplifiers for some forms:

http://www.electronics-manufacturers.com/Optoelectronics/Fiber_optics/Fiber_optic_amplifiers/ [electronic...turers.com]

not sure if that still works for these multiplexed fibres, the optical amplification is narrow band.