Each day, Google’s one million servers, located all over the world, process over one billion search requests, providing more than 7.2 billion page views. The searches come from internet users in more than 180, with queries written in almost 150 different languages; 15% of the queries are entirely new. So what happens when you do a Google search?
1. After a web user types in a question or keywords and hits send, the query travels in packets (each packet, containing a header and a footer, can carry up to 1,500 bytes) amazingly fast — almost the speed of light (186,000 miles per second) to the Google web servers. (Note the sluggish human body is by comparison, where nerve impulses travel at speeds from 2 mph up to 200 miles per hour).
2. The Google web server is a network of more than 3 million computer servers, located in highly protected data centers, that are linked together to form “one brain.”
3. The query is then forwarded to 1 million index servers. When a user does a Google search, he or she is not really searching the entire internet, but rather Google’s extensive index. The Google index is created by internet bots, known as web crawlers of web spiders, that browse the entire internet, visiting each website and following each of its links. As of 2012, Google has indexed more than 50 billion web pages.
4. The query then travels to Google’s document servers that actually do the retrieving of the stored relevant web pages. A proprietary ranking algorithm, known as PageRank (named after founder Larry Page), then evaluates all the data (processing more than 50 million variables and more than 2 billion terms) to arrive at the most relevant pages. In simpler terms, PageRank evaluates how many outside links point to a particular web page, and evaluates how important those links are in order to display the most visited and best known web pages.
5. The Google servers then generate snippets, or short abstracts, that describe or summarize each of the relevant web pages.
6. The information is then delivered to the internet user’s computer.
Elapsed time for all this evaluation and millions of calculations: an astounding .5 of a second!
For further reading: The Human Face of Big Data by Rick Smolan and Jennifer Erwitt, Against All Odds Productions (2012)