Last month's bombshell that the US National Security Agency has been reportedly gathering the phone records of millions of Americans shouldn't come as a surprise to anybody - not least because the technological requirements for such a database are simple and cheap.
In the May 16 issue of broadband e-newsletter DSL Prime, Dave Burstein noted that the type of info the NSA has been collecting - calling party, receiving party, time and duration of call - is only around 30 bytes of data. A database of every call by 200 million people (assuming ten calls a day) for the past year adds up to 23 terabytes. Add a backup database and indexes, and the total storage required is still under 1,000 terabytes. A five-rack server could handle that load easily, and 20 racks could store every call made since 2001, Burstein says.
'I could build it for less than $300,000 as a Linux cluster, mostly out of parts available at my neighborhood CompUSA,' he writes.