Slower performance with ext4

[Rashkae]

Mark Kirkwood wrote:
> Mark Kirkwood wrote:
>> So to reiterate - guaranteeing writes to filesystem is good - but not 
>> good enough if the underlying device does not honor the software 
>> request. This is the guts of most workstation corruption problems, 
>> regardless of fs type.
>>   
> Of course - the above does not rule out implementation bugs with some 
> kernel versions and some driver variants etc etc ... which will be 
> always with us to some extent (c.f the Intel x25-E ssd link in previous 
> posting). We expect (and generally see) these fixed in subsequent versions.
> 
>


It was interesting how the author in that X25 link did not like the
performance with default settings, so first turned off Barriers,, which
exist exactly to prevent the kind of data loss he then goes on to document.

It's all well and good that he then found a way to get great write
performance in a safe manner with write-cache turned off, but I saw
nothing new or interesting in that article