The consequence of Facebook's dependence on MySQL is "worse than death"

Michael Stonebraker, a computer scientist studying database technology, claims that Facebook is running a complicated and huge MySQL implementation, which is equivalent to a "worse fate than death ", the only way out is to repeat all the code. He said the dilemmas of Social giants are common among online entrepreneurs.

According to Stonebraker, Facebook split its MySQL database into 4,000 shards to process the huge data volume of the website. It runs 9000 memcached instances to keep up with the event processing capacity of the database. Facebook has 1800 dedicated MySQL servers and 805 memcached servers.

How does Facebook back up MySQL?

Additional reading]

Facebook's favorite button: Data Gold Mine

Facebook, a social network giant, uses a favorite button and other social agents to record the historical access records of its members and may use the data to locate ads on its own or third-party websites.

According to a Facebook spokesman, the company creates an impression log whenever users log on to a third-party website to view favorite buttons or social extensions, whether or not they interact with them. This record is bound with the unique ID of the user, including the IP address, URL, date, time, and browser information, and will be retained for 90 days.

Facebook insisted that the company did not use these logs to locate ads, and said there is no plan to do so now. However, the extensive use of its plug-ins among most major online publishers has allowed Facebook to construct comprehensive and detailed behavior data. Facebook collects data in a similar way through its plug-in, just as behavior targeting or analyzing vendors placing cookies in web pages or advertisements to track users.

However, behavior targeting companies often compensate publishers for access to user data, but Facebook does not. Say Media Chairman Troy said that it depends on publishers to see if they think the value gained from the plug-in is worth exchanging their viewer data. "Publishers have always faced this problem. Countless manufacturers in the publishing industry use data to exchange value, while Facebook has data and brings traffic. The problem is, in the long run, what if the value of data exceeds the value of the publisher's content or environment ?" He said.

Although Facebook has not yet used the collected behavior information, it is invaluable to advertisers and can bring lucrative profits to the social network. For example, if the social network records the browsing history of a user in ten sneakers-related blogs, a footwear retailer may pay a premium fee, attract that user by serving ads on Facebook.

In another case, Facebook may use data obtained from plug-ins and its own websites to launch its own advertising network, or sell viewer data to publishers, to help them make better use of advertising inventory.

However, a Facebook spokesman reiterated that, without its permission, user data was not sold or transferred to a third party through plug-ins; it does not share users' personal information with advertisers. However, marketers can locate ads based on the user's preferences and interests in personal data, user interaction with favorite buttons, and other plug-ins.