It's not about being big, banks and insurance companies simply have different requirements on how to access and store their data compared to most websites/services.
If you took all of the votes from the 2000 election in the United States (I'll save you the search - it was just over 100 million), it wouldn't even equal one day's ATM transactions in the U.S., and those transactions are a nonevent.
Twitter in not only big in terms of traffic. 200 million tweets per day is not small, even if a tweet involves less processing than an ATM transaction. You can also consider Facebook as an example of website dealing with big data.
http://www.google.com/intl/en/jobs/uslocations/mountain-view...
Google does use MySQL for their AdSense/AdWords transactional system (ie, buying AdWords).
- Oracle and SQL Server: chosen by us
- Sybase and Informix: chosen implicitly by choosing applications that preferred them[1]
- DB2: Chosen by a company that we acquired
Oh and some MySQL and SQLite too.[1] Yes they ran on Oracle too - but when 98% of a vendor's customers are on Informix or whatever, sometimes it's just less painful to go with the flow, so you have access to that community.
It only seems like it isn't when you read sites like Hacker News, where most of the posters are not working in big environments. Facebook is a large scale solution that doesn't use either, but their data management and caching is so bad nobody should be considering them as a best practice.