Cloud computing is quietly changing in Silicon Valley

Cloud computing is not a revolution in Silicon Valley, but a choice, and in many cases it is a choice that has to be made;

One evening in August, the heart of Silicon Valley. In the NASA Research park in Mountain View, California, about 30 people with typical "nerd" (Nerd) temperaments and programmers crowded into a very remote cabin in building 23rd. On the long table next to the hut was a handful of gigantic pizzas, a few bottles of juice and a large bowl of strawberries. The humble little building is home to Carnegie Mellon's Silicon Valley campus.

The room is mostly wearing T-shirts, jeans of young men, do not know each other, the scene appears very quiet, we are just fiddling with their hands of the Apple notebook, individual several also comb the ponytail. The white Apple logo flashed into one.

Just over 6. The diminutive species, Dr. Conco, went to the projection screen and began to introduce to the audience the theme of today's gathering: all the development practices around the GPU (graphics processor) acceleration and HPC (High-performance computing), the presence of experts, academics, front-line developers, There are also engineers from GPU manufacturers who can speak freely about all the topics of interest in the field. The keynote speaker, Dr. Wu Yu, served at the HP Palo Institute, in his "GPU Accelerated Business Intelligence analysis" PPT recommended a reference to the GPU, just introduced a few, the next young man slightly shyly raised his hand: "This book is I wrote."

This is a typical meetup (similar to the salon form of an interest group offline Party). Most of the time, participants who come with ideas, experiences, questions, puzzles and so on can always get something, even a piece of code. Meetup like this are played almost every day in Silicon Valley, with flexible venues and topics that are mostly High-tech, covering virtually any part of the technology circle.

As long as you are interested in a topic, you can easily find an organization on a gathering site like www.meetup.com. For example, the GPU party at Carnegie Mellon's Silicon Valley school building has its organizers and participants from HPC & GPU supercomputing Group of Silicon valley-- A loose group interested in high-performance computing and graphics processors.

Silicon Valley, where the geek and nerd together, technology talk is a wonderful hard currency, a wide range of meetup to a large extent, the "Science and Technology Control" of the social needs of the parties will often appear at the scene of a number of High-tech companies CTO, president and even the founder of the scene, Everyone is good at listening and sharing, and many innovative ideas are brewing in this atmosphere. In the early years, Larry Page, its founder, had been keen to blend in with various meetup before Google's heyday.

Accumulated over the years, the size and activity of groups meetup on the Web site in accordance with technical topics have also objectively become a popular index of some technology. The largest and most active group in the Bay Area,--silicon Valley Cloud Computing Group, a Silicon Valley cloud computing group that was founded on September 11, 2007, is reflecting the attitude of Silicon Valley towards the cloud in 4 years.

Visible clouds

Jack came to the United States from Beijing in the early 90, wrote code and worked as an engineer, and is now an advisor to an investment agency. He's not a geek or a fan of technology, but because of his job needs, he has been a part of the Silicon Valley cloud group's offline gatherings since 2009, almost every game.

According to Jack, in the previous few years, the small size of the group, the general monthly activities, online enrollment, offline came to the scene, the basic is not more than 20 people, because fewer people, the choice of venues is very casual, casually find a university conference room or a café completely get it. by the spring of 2010, the number of online applicants began to rise, and more and more, each party was more and more lively scenes. Each party in June-August attracted nearly 100 listeners.

The meetup boom in the Silicon Valley cloud computing group reached its boiling point in September 2010. As Netflix's cloud computing architect Adrian Cockcroft was invited to share the company's cloud-application experience on Meetup the same month, the number of online registrations quickly broke through 500. Considering the limitations of the site and the difficulty of organizing the work, the Meetup finally accepted only about 300 listeners, who had to arrange the webcast on the spot.

Indeed, Netflix's aura is strong enough to appeal. As the world's largest online film rental service provider, Netflix has been providing the traditional form of DVD-video mailing rental service as its main business since the end of the last century, with a new and efficient membership system and film recommendation algorithm step by step, Its development momentum soon overshadowed the field of the old overlord blockbuster. With the development of bandwidth resources and instant-on-demand technology, Netflix is acutely aware of the market opportunities for streaming media, and has since 2008 started to promote the online transformation of the traditional DVD leasing business, and has opened up a variety of streaming media on demand for its registered members: Netflix's website platform, set-top box, Mobile phones, game consoles and so on, this highly flexible online digital VOD business immediately grabbed a large number of fans. By the beginning of this year, Netflix's membership had expanded rapidly from around 10 million in early 2009 to nearly 25 million. At the same time, Netflix's shares have sparked a frenzy of capital markets, with shares soaring from around $30 trillion in early 2009, breaking 300 dollars this July.

Behind Netflix's saga, however, the company's dramatic cloud trip is the tipping point for the meetup of the Silicon Valley cloud-computing group. Adrian recalls that Netflix's database system crashed on August 11, 2008, leading to a total stagnation in the company's business, and they apologized to users on Twitter's official microblog, explaining that the accident was due to "flaws in the critical components of the hardware system." One side is the key period of the company transformation, must guarantee the stable and unified user experience; On the other hand, the number of registered members and the rapid expansion of business scale, the existing technical support system how to deal with?

Buying servers and building data centers seems to be the only solution, but time and cost? Netflix needs to add more than 1000 servers at a time, and it can take up to 18-24 months to build a full-featured data center, which is expensive, according to industry-prevailing rules. It can be said that the already impatient Netflix is "forced" to choose Amazon's public cloud computing platform. The low-cost, highly flexible Amazon public cloud service is the only viable option for Netflix's current situation.

The journey of the cloud is not all joy. Traditional enterprise software can not be smoothly migrated to the cloud, the most fundamental reason is that the traditional software development process, hardware and software is a very tight coupling state, and in the cloud, all is loose coupling. Netflix's solution is that they have locked 12 of the company's core developers in a room for 2 days, forcing them to forget traditional development thinking, and to rewrite enterprise apps in a cloud. Later the practice proved that they succeeded.

Change has taken place. Adrian said a year ago that nearly 80% of Netflix's online video leasing business was done entirely on Amazon's cloud computing platform, with only a handful of very core business materials, such as membership information, still in its own data center. Netflix hopes to reach 100% of its cloud platform by 2014 and become a cloud-driven company.

Is it reliable to put your key business into someone else's data center? Kevin McEntee, vice president of Netflix, replied that Amazon's redundant architecture of tens of thousands of servers is clearly more reliable than the company's own data center. The company's research team can instead devote more effort to improving the experience for the vast number of users.

Users don't care where your data center is, but just what you're offering. More importantly, the cloud is almost perfect for companies with a sharp "burst" of power like Netflix, because of the high flexibility and scalability of the cloud, which Netflix can do to achieve explosive growth.