Google Daniel Jeff Dean is how to become the Internet Ares

"The speed of light in a vacuum was about 35 miles per hour, and Jeff Dean spent a weekend optimizing basic physics. "--from" facts about Jeff Dean "

In fact, the "facts about Jeff Dean", as described in this g+ post, are not true. But it was unusual for him to build a website similar to "The fact about Chuck Norris." This is because Jeff Dean is a software engineer, and software engineers are usually not like the martial arts champ Chuck Norris. On the one hand, they are not Lone Ranger, software development is essentially a collaborative process. On the other hand, they never use submachine guns to hit Cowboys like Chuck Norris in this video.

Chuck Norris (Chuck Norris) is a karate world champion and an American film actor. He has another more well-known translation "Rollis", from the Kung Fu business card "dragon over the river." He developed the film career at the beginning, in Bruce Lee's martial arts film "Dragon Across the River" played a karate master colt, and Bruce Lee in the Colosseum Duel, is recognized as the classic martial arts scene. (Excerpt from Wikipedia)

Jeff Dean

However, on April Fools ' Day in 2007, some of the young Google engineers who came to see it were supposed to give Jeff Dean a website to compliment him on his programming achievements (translator: The following are just jokes, not really). For example:

The compiler won't warn Jeff Dean, and Jeff Dean will warn the compiler. Jeff Dean writes the binary code directly, and then he writes the source code as a document for other developers to read. When Jeff Dean was thinking about ergonomics, it was to protect his keyboard. One day when Jeff Dean was optimizing a feature, he was forced to invent an asynchronous calling API. In this case, the function can return the result before it is called.

A real truth about Jeff Dean is that you have to be a computer whiz to understand a lot of the jokes that people say about Jeff Dean. (interested readers, the Business Insider website offers some explanations of his more popular jokes.) If you don't have a background in computer science, it's hard to understand what those jokes say about his false accomplishments, not to mention the real achievements he's made at work. Dean's own systems, such as MapReduce, BigTable, spanner, and so on, are not known to many Google users from Goolge. However, these programs are the cornerstone of Google's and modern internet presence. Some of the projects he is currently working on are likely to bring about a revolution in information technology again.

When you think about who created the Internet today, you might think of the founders and CEOs of many companies, such as Tim Berners – Lee (Tim Berners-lee), Mark Anderson (Marc Andreessen), Larry Page and Sergei Brin (Sergey Brin), perhaps Mark Zuckerberg. It makes sense that every one of these people has invented a product or framework that shapes the way we use the Internet today.

At the same time, in the shadow of these giants who have been freed from their daily work, it is a bunch of unknown developers who, on the keyboard, knock out every day to give us the products and systems we use. Unlike other industries, these people are often irreplaceable in high-tech industries. A good accountant may save you 5% of your personal income tax. A good baseball player's upper base rate is a little bit higher than the average player. (translator: Baseball is generally used to the success rate, batting success rate, etc. to measure the player's level). But it may take a team of nearly 10 people a few months to complete a good software developer's work during the week. This difference is at the exponential level. This is not a fact about Jeff Dean, but the knowledge of Silicon Valley's high-tech industry, which is why the best companies spend so much on attracting top talent.

When Dean joined Google in 1999, he had the honor of being the top young computer scientist in the United States. At the beginning of the popularity of home computers, Dean said he was always looking for ways to keep track of extreme performance on a given machine. When he was a high school student, he wrote a software to analyze a large number of epidemiological data. He says his software is 26 times times faster than professional software at the time. This system, called Epi Info, is adopted and translated into 13 languages by the Centers for Disease Control (Centers for Disease Control). When he was reading a PhD in computer science, he studied the compiler, which was used to translate the program source code into a language that the computer could execute. "I always like code that runs fast," he says.

But Dean is not a complacent person, he does not want to spend his life on the compiler, so he later left the academia. In less than three years, he joined Google, which was only 20 people at the time. (according to Steven Levy's "in the Plex", Google, as a search-start-up company, thought Dean was a hard-won talent.) He made a significant contribution to early Google News and AdSense, where AdSense, the advertising product, rewrote the rules of the Internet company's game. After that, he shifted to focus on one of the company's core issues: extensibility.

The original idea of Google's basic algorithm came from Page and Brin, both of which were top developers at the time. In the late 90, they created the PageRank algorithm, an algorithm that returns the most relevant search results when a user is given a query. The focus on the relevance of search results has made Google go beyond Yahoo, AltaVista and other search engines that were in the lead at the time. But as Google has become more successful, it has encountered a huge technical challenge. "We can't deploy more machines quickly enough to respond to demand," Dean recalls.

So Dean and his colleagues, including another great programmer, Sanjay Ghemawat, found a solution. The problem, as he did with Epi Info in high school, looks like a hardware problem. Ghemawat helped lead a team to develop the Google file system, GFS, which allows very large files to be distributed across a wide range of inexpensive servers. Dean and Ghemawat, together, developed a programming tool called MapReduce to help developers effectively use these machines to process large datasets in parallel. Just as the compiler helps programmers to write programs without thinking about how the CPU handles the program, MapReduce allows Google's developers to tweak the search algorithm or add new functionality without worrying about how to parallelize the operations or how to handle hardware failures.

Dean and Ghemawat's approach was so powerful that when they published a research paper at a conference in 2004, the method immediately became the industry standard. Until today, MapReduce has become an important cornerstone of many other projects, one of which is the famous open source framework Hadoop. And it is Hadoop that has created the new buzzword "big data" in the industry. Big data is used in different areas, from online travel to energy exploration. And just as Google began to expand from MapReduce to other new programming models in some core applications, Dean said he would start using mapreduce a lot when he saw many summer interns starting new projects after Google.

MapReduce is a good example of the 10 times-fold effect that one of Google's founders page says. The 10 times-fold effect is about 10 times times better than the original one, not 10% better than the original. MapReduce does not make a particular type of operation a bit faster, but it helps every developer in Google do what they couldn't have done before.

Some of Dean's other projects also have a similar exponential effect.　On the basis of Google's file system, he and Ghemawat created a distributed database system called BigTable. The bigtable can handle 1PB of data (1 pb=1 million GB. (Translator: The Open source community also has a similar project, based on the Hive on top of Hadoop) after which they further developed the spanner system called the world's largest single database. The Wired case Metz says that by using innovative time synchronization, spanner's physical storage spans the world's different data centers, but operates like a place. In other words, it enables different information in the global data center to be consistent, even if a particular update request may take a different time to reach different data centers. Metz also said that before Spanner was reported, no one ever felt that the system could be built.

Now it seems that the real truth about Jeff Dean looks a bit like a fake. Dean himself would have laughed at the situation, saying it was awkward, but also a compliment. But, he added, it is to be remembered that the true accomplishments of his work are always achieved through co-operation with different people.

Almost every morning, he goes to work at Google headquarters in California State, Mountain view, and always sits down and drinks coffee with the same people. He estimates that over the years we've probably wiped out 20,000 cups of cappuccino together. These people don't always work together. In fact, some people have moved to different offices on the other side of the Google campus. But when they get together and talk about what they're doing, some people's problems can always inspire other people's new ideas. These coffee chats have led Dean to apply his experience in optimization, parallel computing, software architecture and many different types of projects. This made him have enough ambition and self-confidence. "He's always been passionate and optimistic about what we can do," says Ghemawat, a longtime partner, "and nothing can stand in the way."

His recent work can be a good indication of what Google is going to do next. Last year, he worked with Andrew Ng, an expert in machine learning at Stanford University and one of Coursera's founders, to help Ng's postgraduate Quoc Le perform an unprecedented test of unsupervised machine learning. The experiment, under Google's secret Google Skunk Project (the translator: a Secret Innovation project), uses 16,000 of processors to learn about the YouTube video without doing it, coming up with a way to identify a cat. This seems to use a lot of computers to come up with a very basic result, but this experiment can help us lay the groundwork for the next generation of AI technology. Future AI technologies will play a role in many potential applications, including Google Now, which uses personal assistant technology, and image search, which is a great help for the Google Eyewear project Glass.

Jeff Dean may be inventing something incredible, just like the "facts about Jeff Dean" mentioned in the 0 and 1 special keyboards (translator: This keyboard does not exist, this page of the article is joking that Jeff Dean is directly using binary machine code to write programs. Jeff Dean admits that he is not an expert in machine learning, but he is happy to use it to help with the research in building scalable, highly available systems.

Contrary to the "facts about Jeff Dean", Dean says the best way to solve problems in many situations is not simply to sit down and start writing programs. His approach always starts with some simple calculations to find the best balance between quality and speed for a particular process. In many areas, from machine translation to search quality, you always try to weigh the amount of computing you can do with each query, he says. You may not be able to find the ideal solution, but we can always get 98% of the benefits through a 1% calculation in some approximate way.

Dean often does this kind of calculation so that he gives a "list of numbers that every computer engineer should know". These include, for example, how many milliseconds (150 milliseconds) to send a network packet from California to Amsterdam at the speed of light. Keep these numbers in mind, and in 20 minutes you'll be able to differentiate from the whiteboard which one of the 3 designs will be the best. And he said, if you can't do it quickly? It would be easier to multiply by converting all of these numbers approximately to 2 of the time Square.

If Dean really has the power of Superman, then this ability does not make things perfect in an instant. This is the ability to weigh, optimize, and deal with problems at different levels of things. In other words, it is a ability to discover opportunities and to make things as good as possible in a short time, rather than trying to be perfect at the beginning. In Silicon Valley, this is much cooler than taking a submachine gun to the Cowboys.

Google Daniel Jeff Dean is how to become the Internet Ares