Over a decade ago, there was a lot of hype that telcos would take over the payment industry ... that their phone call transaction systems were positioned for handling the scaleup (that would happen with things like micropayments, cellphones, etc).

A couple yrs later (when it wasn't happening), the excuses were that the telcos weren't positioned (didn't have enuf infrastructure, experience, etc) to handle the financial liability (and fraud).

recent post mentioning the scenario
http://www.garlic.com/~lynn/aadsm27.htm#66

in that timeframe there were some number of "in-core" dbms systems developed; they supported transaction semantics ... but the default location for records were in memory ... as opposed to the default record location on disk (and memory used for caching). these products claimed something like ten times performance improvement vis-a-vis traditional disk-oriented dbms implementations (using same exact configuration and real storage size) ... and found lots of uptake among telco customers. In the past couple yrs, there has been announcements about some of the payment networks installing these products (usually associated with comments about supporting scaleup issues).

some old posts:
http://www.garlic.com/~lynn/2004e.html#25
http://www.garlic.com/~lynn/2007m.html#47

for topic drift ... some comments about early uptake of relational dbms coming with the increases in system real storage sizes (caching information used by rdbms to automatically handle stuff that required manual administration and application programming in earlier dbms implementations)
http://www.garlic.com/~lynn/2008c.html#78
http://www.garlic.com/~lynn/2008c.html#88
http://www.garlic.com/~lynn/2008d.html#11