Google once faced a fatal problem that prevented it from creating a search index.
When you want to know something or access a page, just enter the phrase you want to search for in your browser and it will instantly return the latest results. Information search is carried out using a complex mechanism that incorporates concepts from various academic disciplines called
The Friendship That Made Google Huge | The New Yorker
https://www.newyorker.com/magazine/2018/12/10/the-friendship-that-made-google-huge
Google's core system, which crawls the web and creates a search engine index, stopped in October 1999. Users could still search on Google, but if they searched for something in March 2000, the results were stuck at five months old, and they had problems accessing the latest news and web pages from the search results. In addition to users, there were other important issues. At the time, Google had a contract to provide a search engine index to run Yahoo!'s search engine, and if the system stopped completely, the contract would collapse and Google would even be in danger of running out of funds and disappearing.
Around March 2000, a group of system engineers led by Craig Silverstein, a software engineer at Google's founding, organized an intensive troubleshooting meeting. However, after spending four full days reviewing the code, they were unable to come up with any improvements. Silverstein later recalled, 'None of the analysis we were doing made sense. Everything was broken, and we didn't know why.'
Jeff Dean , who joined Google from Digital Equipment Corporation (DEC), one of America's leading computer companies, in early 1999, and Sanjay Ghemawat , who joined Google from DEC in December 1999, also dealt with system problems during their short careers at Google. Dean later became head of Google's AI division.
Dean and Ghemawat had been good friends since their days at DEC, and they liked to work side by side writing code together. They spent days digging into the stopped index, finding code defects such as missing words and incorrect order, but could not find any obvious bugs. On the fifth day of the project, the two began to suspect that the problem was not software, but physical. They tried to understand what the machine was looking at by simplifying the messy and cumbersome index file into binary code represented by binary numbers.
In binary code, a monitor displays a string of 1s and 0s. Each string represents data in the index, but at one point, Ghemawat noticed that a number that should have been 0 was a 1. When Dean and Ghemawat put together all the words that were misprinted, a pattern emerged. As a result, it turned out that the machine's memory chips had somehow become corrupted.
Google was founded in 1998, and from 1999 to 2000, when the trouble occurred, it was still being run like a startup, The New Yorker points out. In fact, there were failures and malfunctions in various aspects, such as motherboards, hard drives, and wiring. NASA and financial institutions used special hardware that would not affect the whole even if a single bit of '0' and '1' was flipped due to a hardware malfunction, but Google used cheap computers and could not avoid hardware failures. As a result, Dean and Ghemawat wrote code to compensate for the problematic machines, and a new index was completed, and the countermeasures meeting was dissolved.
The New Yorker points out that the root cause of this problem is the origins of Google. Google's founders
At the time, Google had a stack of 1,500 devices that combined motherboards and hard drives, but only about 1,200 of them worked because of hardware failures. Faults caused by hardware failures continued to destroy the system, so Google needed to integrate the computers into a resilient whole. To prevent the problem from recurring, Dean and Ghemawat wrote code so that the entire system would not go down if one hard drive failed. Alan Eustace, a computer scientist who became head of the engineering team in 2005, praised the two men's abilities, saying, 'Paradoxically, to solve large-scale problems, you need to know the details. Dean and Ghemawat understood computers at the bit level.'
Google engineers are ranked in stages, with level 1 being IT support staff, level 2 being fresh out of college, and level 3 being master's degree holders. Level 4 requires several years of contribution or a PhD, and most staff stop at level 5 even after a career. Level 6 engineers are the top 10 percent, level 7 is level 6 with a long track record, level 8 is for those involved in major products or infrastructure, and level 9 is for well-known engineers who are spoken of with respect. Level 10 is a world-leading, influential, and expert executive. And Dean and Ghemawat are senior fellows at Google, the first and only two people at Google to be certified as 'level 11'.
Dean and Ghemawat prefer ' pair programming ,' where two people work on one computer. Ghemawat said, 'I don't know why we decided on pair programming. If you can find someone who is compatible with your way of thinking, you two complement each other. I don't know why more people don't do it now.' It's rare to find a software engineer with such a good partnership as Dean and Ghemawat, and Dean and Ghemawat are closely connected not only at work but also in their private lives, including their families. The New Yorker described the incident in which the two of them solved the problem as 'the friendship that made Google huge.'
Related Posts:
in Software, Web Service, Hardware, Posted by log1e_dh