Re: wikipedia down, slashdot covering, mysql mentioned

[DebugasRu]

>> Many data centers do not allow customers to install their own UPS
>> inside the rack.

The questions is not whether they have (wish to have, can afford etc ) UPS or not,
or if the OS is buggy ( fsync() function or others ),
the problem is the database management system failed to ensure
data integrity under stressed conditions. Of cause ensuring persistance of transactions
when
underlying hardware/software is unreliable is a very difficult task requiring
multiple trade-offs and that's why i find it impossible to compare
performance of two given DBMS where one of them ensures data integrity
and the other one cannot. What would you prefer - the system that
fails in 1/billion's occasions but runs 10 times slower or the system
that fails 1/100 occasions even though it runs 10 times faster? Or is
it better to have a system that never fails but runs 100 times slower?

> Probably fsync() had failed to flush some part of a 16 kB page to disk.
so what ? one of trade-offs would be to re-read the data from the disk
and compare it with what it should be (another copy on the disk) and only after that fix
the
transaction, otherwise roll it all back.
When you have multiple users keeping db copy for each of them
(versioning) becomes a nightmare...
Solutions do exist but they require expensive trade-offs and more
complex algorithms.