Re: In-memory database

On Mon, 28 Oct 2024 23:45:53 -0000 (UTC), Waldek Hebisch wrote:

I think that transaction center wants to keep more information
than a single copy of data in the RAM: with single copy any
memory corruption could mean loss of hours of transaction data
which is equivalent to quite a lot of cash.  So I suspect that
that there are layers of redundancy buit-in.

They could distribute the load, by spreading it across multiple processing
centres. For example, most transactions on a given card are likely to be
with businesses in a particular locality, or with certain large online
retailers. The card has a credit limit, but I suspect that is not a “brick
wall” limit, so if it takes a few seconds to reconcile multiple
transactions, with the chance that they could add up to something a bit
beyond the credit limit once the totals have been made consistent again,
that’s not the end of the world.