[ltp] Re: solid state drive?

[Laurent]

Hi,

>> The journal causes a lot of small writes.  Small writes have a bigger
>> write amplification, so it's a lot more data for the SSD. That makes
>> it slow.  I don't worry about the number of writes, but i do want
>> a fast system.
>
> So that which makes you say to avoid ext3 is your expectation of how you
> think it works, which doesn't actually correspond to how it does work.
> Right?

For my workload (lot's of syncs) it does work the way i described it.
The journal is updated at each sync (=after each write).

> Or have you actually measured the performance problem?

ext3 is 3:1 slower then ext2. For my workload.

cu