2016年2月29日 星期一

Google 的高空網路氣球計畫正式啟動

目前 Google 已經在波多黎各完成了氣球的測試工作,首個正式發射升空的氣球將在印度尼西亞升空,2016 年年內完成。Google 公司表示 2016 年將在全球各地設置高空網路氣球的站點,構成一個完整的網路接入體系。

Google 的高空網路氣球計畫正式啟動,不過相當於一個網球場大小的氣球該如何讓它飛起來呢?他又是如何提供全覆蓋的無線網路服務?

Google 高空網路氣球 Project Loon 即將正式推出,該公司首次曝光了氣球升空的過程,計劃在 2016 年內建立一系列站點,高空氣球可在空中停留 100 天。 Google 高空網路氣球相當於一個網球場大小,氣球升空需要一個大約…

2016年2月28日 星期日

Global Supply Chain by Amazon. The Pros and Cons of Amazon Buying FedEx

近幾周盛傳亞馬遜(Amazon.com)計畫開創全球物流業務,與UPS和聯邦快遞(FedEx)競爭。在1月28日的財報電話會議上,有人就亞馬遜租用飛機、註冊海運公司等自建物流的報導詢問了首席財務官奧爾薩夫斯基(Brian Olsavsky),他說那是要在物流旺季作為其他公司的補充,不是要取代它們。
亞馬遜的內部文件揭示的計畫則比這個大膽的多。一份在高階主管層傳閱的2013年報告提出,亞馬遜配送服務(FBA)可以進行大規模的全球擴張,這項服務會向透過亞馬遜出售商品的獨立商戶提供倉儲、包裝和物流。報告勾畫了一張全球物流網路,可以控制從中國和印度的工廠到亞特蘭大、紐約、倫敦顧客家門口的貨物流向全程。這項被稱為「龍舟」(Dragon Boat)的計畫正在快馬加鞭地進行。
內部文件稱,亞馬遜最快在今年就會創立一個叫做亞馬遜全球供應鏈(Global Supply Chain by Amazon)的事業專案。它的目標是將亞馬遜置於一個物流產業的中央,這個產業過去包括聯邦快遞、UPS等託運商,以及眾多負責跨國貨運和手續的中間商。一旦繞過這些中間人,亞馬遜可以把全球成千上萬的商戶的貨品匯聚到一起,以更低的價格購買卡車、飛機和貨船的空間。商戶可以在網上一鍵索取貨櫃空間。
「這是經典的亞馬遜風格,」金融服務公司貝雅(Robert W. Baird)分析師塞巴斯蒂安(Colin Sebastian)說,他認為公司的物流業務可能成為一個價值4000億美元的生意。「他們在用很小的步幅走一條漫長的路,這就讓一些可能會被顛覆的公司繼續拒絕面對現實。亞馬遜很少一下子走出一大步,給市場帶來震動。」
方案提出三年後,有跡象表明該計畫即將成為現實。一月,三藩市物流公司Flexport在一篇網誌中指出,亞馬遜的中國分公司已經在美國註冊,向其他公司提供海運服務。一月前,一名瞭解計畫內情的人說,亞馬遜考慮租用20架波音767運輸機,以進一步控制其物流成本。「有一點很明白,」電商分析公司Stella Service行政總裁萊瑟(Jordy Leiser)說,「亞馬遜希望把它的優勢轉化為錢。」——Spencer Soper;譯 經雷

Is it totally crazy or worth considering?

The e-commerce giant may have better options for building out a logistics network.

Amazon's aspirations of building a global logistics operation have spurred rumblings about whether a takeover of FedEx may be a cheaper and swifter means to an end. Is there any merit to the idea?
At the very least, it's an interesting thought experiment.
First, let's look at the pros. In an effort to curtail rising shipping costs, Amazon has built a bevy of distribution centers and tested alternative delivery methods. The $245 billion e-commerce giant reportedly wants to turn those logistics baby steps into a full-blown delivery network that can ferry goods from China to New York. Buying FedEx would certainly get Amazon a ready-made global operation in a hurry.


FedEx had 656 planes in its Express division as of November and uses about 200,000 vehicles and trailers across its network. It also pays some 300,000 employees to make the whole thing work. Building up that kind of system would take at least a decade, if not much longer. For one thing, Boeing and other aircraft suppliers don't just have 500 planes sitting around for the taking.
Whether it's cheaper or not depends on how big you think Amazon's logistics ambitions are. If it wanted to build a network identical to FedEx, we're talking at least tens of billions of dollars for the planes alone. Say Amazon wanted to start out with about 180 medium-body freighter planes and 20 wide-body freighters for bigger trips. Even taking into account some degree of a discount for buying in bulk, a plane order of that size could run Amazon in the range of $12 billion, according to estimates from Kevin Sterling of BB&T. And that would give it just a third of FedEx's air fleet without any trucks, real estate or additional employees.

Once you start tabulating that all up, maybe it is easier to just buy FedEx. Any buyer would probably have to pay a premium to the company's all-time high in June of $184.98. That puts a deal in the range of $60 billion, including debt. It's not chump change, but Amazon has about a third of that in cash and equivalents already. It's at least conceivable financially.

Packaged With A Premium
FedEx has slumped amid a few quarters of underwhelming earnings, but any buyer would likely have to pay a premium to its June high.
Source: Bloomberg

Now for the cons. Amazon doesn't really want to be FedEx. It wants to control its own (smaller) version of FedEx. In that respect, a deal really wouldn't be all that cheap.

There's an obvious benefit to Amazon in taking control over the delivery of goods sold on its site and those of smaller merchants with which it has a relationship. There's far less appeal in ferrying a package from my mom in Kansas to me in New York, for example. As noted here, it's also highly unlikely that major merchants such as Wal-Mart that compete with Amazon would want to use the e-commerce retailer for shipping services.

Of course, Amazon could just leave these types of deliveries to other providers, but then why would it go to the trouble of paying for all those capabilities in a FedEx acquisition? The point is, FedEx comes with some extra baggage that Amazon probably doesn't want or wouldn't be able to use to its advantage. That's not worth a premium. 
Amazon still has work to do to build out a global logistics operation, but it's already taken those baby steps. It has more than 100 fulfillment facilities globally, sorting centers that function as hubs and an intercity delivery network in some areas. It could close the gaps in its current network for a quarter of what it would cost to buy FedEx, says Satish Jindel, a logistics consultant.
One way to do that could be through a series of smaller deals for regional carriers. Amazon already works with a number of these providers, which can offer faster and cheaper delivery than the national providers within their respective areas. Buying a few of these regional operators, such as closely held LaserShip Inc., would help give Amazon more control of the so-called "last mile" aspect of its delivery operations that it currently outsources -- without a $50 billion-plus tab.
Another option may be Air Transport Services Group. The company has a fleet of the Boeing 767 freighters that Amazon has reportedly expressed an interest in and Air Transport is adding flights for an unnamed customer widely believed to be Amazon. With a market value of about $740 million, it's a bargain way to add air transport capabilities. 
That's more Amazon's style anyway. The biggest deal the company has done -- and its only one worth more than $1 billion -- was a real estate purchase. Amazon's preferred method is to work with and learn from partners, and then run them out of town (see: Borders). All signs point to it trying to do a similar thing in the shipping industry. 
This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
  1. Actual prices are difficult to determine because buyers get a discount to the catalog price. The price may be lower than estimated if the discounts are more substantial.  

2016年2月25日 星期四

 Google's new quantum computer is '100 million times faster

Google's new quantum computer is '100 million times faster ...

www.telegraph.co.uk › Technology › Technology News
Dec 9, 2015 - Google and Nasa have been working on a lightning-fast quantum computer that is 3,600 times faster than a supercomputer at solving complex problems. ... Has Google won the race to build the world’s first commercial quantum computer? ... Google and Nasa announced they were collaborating on ...

2016年2月24日 星期三

 Facebook Totally Redesigned Like Button, 用人工智慧建立世界人口地圖,幫無人機找出未連網地區

In the news
Image for the news result
"Like" you already know— now say hello to "love," "haha," "wow," "sad," and "angry".
Reactions Now Available Globally
Facebook Newsroom - 14 hours ago

撰文者:郭芝榕 發表日期:2016/02/24
Facebook連網實驗室(Connectivity Lab)決定用人工智慧技術解決這個問題,推出一份精確的全球人口分布地圖,使用高解析度衛星圖片,並用深度學習來標出圖片中的每個房子,找出人口分布在哪些區域之後,再選擇要用哪種方式來發送網路,將會更有效率。




Facebook連網實驗室總監馬奎爾(Yael Maguire)說,目前世界上最好的人口資料是哥倫比亞大學,但也只能看到1公里解析度的資料,仍然不夠詳細。「我們想知道城市以外的地區的人口分布,但資訊非常不足。」
於是一年前,連網實驗室跟Facebook的人工智慧、資料科學部門,以及第三方衛星資料單位DigitalGlobe’s Geospatial Big Data合作,找出20個國家過去五年的高解析度圖像資料(每像素50公分),覆蓋約2160萬平方公里,資料量共350TB,人工智慧團隊處理了146億筆圖像資料。最後用深度學習演算法教會機器如何從圖片中辨別人類居住的區域,分析的精準度超過90%,希望能建立5公尺解析度的圖像資料。
  • 劃分出30公尺乘以30公尺的區域,當作候選區域,此舉可以先排除包括水和沙漠等無人的地區。
  • 再使用Facebook的圖像辨識引擎和深度學習演算法卷積深經網絡(Convolutional Neural Network, CNN),找出圖像的特徵。
  • 人類被標記在候選區域的一小部分,為了訓練和優化在不同的地理區域的各種類別,建立模型,在衛星圖像中找出人類所居住的區域。

2016年2月22日 星期一

Facebook's Huge Plan; Telco Infra Project, or TIP

Facebook's Huge Plan to Shake Up Dull World of Telecoms

Stacey Higginbotham
@gigastacey FEBRUARY 22, 2016, 2:00 AM EST

Cisco isn’t going to like this.

Facebook has teamed up with Nokia, Intel, and several telcos to create an open source hardware and software platform for the telecommunication industry designed to make it easier and faster to build communications networks. The program is modeled after a similar effort the social network created five years ago for its data center.
If successful, the project will help operators build networks using a more modular approach, but it will also change the economics of telco equipment providers such as Ericsson and Cisco   .
Facebook  FB 2.48%  is proposing the Telco Infra Project, or TIP. The goal is to build open source hardware and software that will cover the three elements of a telecommunications network: The access, the backhaul and the core. Access is how a phone or a computer gets online in the first place and includes elements like the base stations and cell towers. The backhaul is how the signals from the devices get onto the Internet itself, and includes fiber access or even copper or microwave, depending on the situation. Once it gets to the Internet, those data packets have to get where they’re going and be counted appropriately on the cellular network for billing or application purposes (things like traditional text messages are still routed differently than WhatsApp messages). This happens in the core.
Get Data SheetFortune’s technology newsletter.
That’s telecom 101, and, because it’s a system that has been built up over a century, it’s a mish-mash of complicated gear and proprietary software holding it all together. However, the Internet, and particularly the evolution of services that go over the top of carrier networks using data connections, have threatened the carrier’s business models. It’s hard to charge people 20 cents per text, for example, when Facebook Messenger or WhatsApp are “free” with a data plan.
The proprietary, cumbersome, and expensive mess of gear is holding carriers back. It raises their costs, keeps them from innovating quickly and forces their executives to fear becoming a dumb pipe.

Enter the cloud model

The solution for carriers has been to turn to cloud providers for inspiration. The big question for the telecommunications carriers is whether they can turn fast enough—and whether their vendors will move with them. For example, this shift in underlying infrastructure is behind AT&T’s move to adopt software-defined infrastructure. That infrastructure will hopefully give Ma Bell more flexibility to deliver services without having to manually change out its gear, and also helps it simplify the overall network design. As part of AT&T’s  T 0.79% plans to build this new type of network, it announced last week that its SDN efforts were 5.7% complete in 2015 and would be 30% complete by the end of 2016. It plans to roll out its next-generation wireless network using equipment from Ericsson and Intel.
For more on AT&T’s plans for software-defined networking watch our video:
Facebook’s TIP has a similar goal to AT&T and is even working with Intel  INTC 2.19% . Other partners include Deutsche Telekom, Nokia  NOK 3.17% , SKTelecom and Globe Telecom.
Facebook hopes to create a more simplified network architecture using modular components that combine hardware and software. This would replace the expensive, proprietary gear that underlies telecommunications networks today. Facebook calls the efforts open, but there will be licenses involved, simply because the telecommunications world is far more complicated than the world of building servers, where Facebook has experimented with open source hardware before.
In April 2011, Facebook launched the Open Compute Foundation, an effort to build open source servers. At the time the goal was to help eliminate much of the unnecessary stuff that the server makers put on servers that Facebook felt added costs in terms of components and power draw. For a company like Facebook, which had tens of thousands of severs, $2 in extra components costs added up. A year ago, the company said investing in open source servers had saved it more than $2 billion.
What was good for Facebook and other big buyers of server gear was less of a win for the makers of the boxes themselves. While Dell and HP  HPE 0.95%  were on board as Open Compute partners from the get go, it’s no secret that the margins of building bare bones servers for the most demanding buyers are slim. Dell has made a go of it, while HP has struggled. As Facebook tries to bring this model to the telecommunications world, which has seen its business hurt by consolidation as well as the recent move to more open systems, things are about to become brutal.
So far, of the equipment vendors only Nokia is a member of TIP, leaving out the two largest telco equipment vendors Ericsson  ERIC 1.08%  and Huawei. Ericsson did just join the Open Compute Foundation, the open source server project, which also counts Nokia as a member. Nokia has less to lose by becoming involved in any effort to build modular telecom gear. That was already a move that Alcatel-Lucent, the company that Nokia merged with, was veering toward. However, for other providers of gear, who tend to sell their equipment as a highly integrated system, modularity represents a threat.
Being able to pick and choose from a variety of vendors creates competition for the various elements of the network, where it didn’t really exist before. Axel Clauberg, a VP of aggregation, transport, IP & fixed access at Deutsche Telekom AG, said via email he’s looking forward to the lowered costs and potential boost in efficiency.
We believe that the exponential growth of Internet traffic requires new approaches, also on networking equipment like routers and switches. The Open Compute Project has proven that open specifications for hardware, combined with an active community, can have a drastic impact on efficiency and cost—TIP will trigger the same for all areas of the network.
As for existing vendors, Clauberg says of including companies such as Cisco and Juniper  JNPR 0.85% (whose gear is also likely to be hard hit by this initiative), “We are expecting similar changes as they already happened within the data center. While this impacts existing players, it also creates large opportunities.”

TIP in action

So far Facebook working with Globe has created the first pilot network using the TIP gear in a small village in the Philippines, and Jason Taylor, VP of infrastructure at Facebook, estimates that more networks will be tested later this year. Carriers will likely reap the benefits of having this level of modular design first, although the savings may not be as drastic as what happened in the data center because of the level of complexity that telecommunications infrastructure requires.
The equipment vendors that move quickly to reduce their costs and adapt to build products for the TIP will have an easier time than those who ignore it as a flash in the pan or something that won’t have much an impact. One question worth pondering is how far the TIP efforts will extend. When Facebook introduced the Open Compute Project, it was mostly adopted by the companies that operated giant data centers and those in the financial service industry. Server makers still could count on enterprise clientele to buy their fancier boxes at higher margins. But the telco equipment market has a much smaller group of buyers.
So it’s not clear if there will eventually be a bifurcation of buyers who will adopt the Facebook’s Telco Infra Project while others steer clear. Even those who don’t play along are likely to adopt the principles of modularity and flexibility through software-defined architectures and buying gear from companies that support true interoperability. That’s why vendors like Cisco and others have been leery about the threat of open source switches, Open Flow, and other elements that have been forcing the network world to open up and embrace interoperability between vendors and their gear.
With TIP, Facebook is basically creating a framework and consortium to ensure that interoperability happens without all of the FUD and games that typically happens in an industry that is long overdue for a change. The fact that Facebook has convinced Nokia and several carriers to get involved is a great first step and one that makes TIP worth watching. If I were Cisco, Ericsson, Huawei or Juniper, I’d be looking over my shoulder.
Photograph by Chris Ratcliffe/Bloomberg - Getty Images