R
Roedy Green
In-RAM databases are orders of magnitude faster than disk based SQL
engines. RAM is getting cheap enough that many databases, or at least
the crucial part could be totally RAM-resident.
You then have two problems:
1. handling failures
2. persisting/exporting the data.
You could handle failures by having paired machines each running off a
UPS running in tandem.
you could handle the persistence problem by some sort of POD or export
that runs in the background cycling through the objects, perhaps
several minutes behind the changes, or by logging transactions that
could later be replayed.
Whatever your solution, you bundle it up so that it becomes as
convenient to use as a POD or SQL database.
engines. RAM is getting cheap enough that many databases, or at least
the crucial part could be totally RAM-resident.
You then have two problems:
1. handling failures
2. persisting/exporting the data.
You could handle failures by having paired machines each running off a
UPS running in tandem.
you could handle the persistence problem by some sort of POD or export
that runs in the background cycling through the objects, perhaps
several minutes behind the changes, or by logging transactions that
could later be replayed.
Whatever your solution, you bundle it up so that it becomes as
convenient to use as a POD or SQL database.