Question for the Commentards!
RE a related bank cockup, and some possible future ones:
HelloSailorBC Global Retail in London started having their ingathering-of-all-globaloffices daily upload crash when its size grew too large (over 10gb IIRC), about 3-4yrs ago. Someone "fixed" it in 2secs by changing the file-load method (in C# IIRC) from normal to the memory-mapped one. This SHOULD have had no effect or actually worsened it, but instead solved the problem immediately.
What this implies is two-fold:
1. C#'s(?) file access library uses a Schlemiel-the-Painter algorithm on normal reads, AND
2. there's a bug in the low-level Disk Driver or Storage Driver: use it too hard and it falls over (suggests a race condition)
The latter is what interests me. Anyone know what setup that company uses, under the hood, for their storage?
Because it's buggy.