[clue-talk] Benefits of SSDs : was new processors

[NICK VERBECK]

This may be covered in the links that have been provided, but in
regards to the 1st email. If you can avoid it don't mount your swap to
an SSD. Yes you will receive better speeds, but at the cost of the
life of your drive. By the nature of how SSD work, the constant
writing & rewriting of data to the drive degrades the life of the
drive as well as speed overtime. There have been firmware updates to
improve the life and speeds of the drives, but it is still suggested
to reduce that kind of work if you can.

The best solutions I've come across is to use the SSD for you OS
partions and use a SATA for everything else. You will still get you
decreased boot times and application speed improvements as that stuff
will continue to live off the SSD. It will also save you a bit of
money while we wait for SSD costs to come down as you wont need such a
massive drive anymore.

On Mon, Jan 4, 2010 at 9:43 PM, Maxwell Spangler
<maxlists at maxwellspangler.com> wrote:
> On Mon, 2010-01-04 at 16:49 -0700, chris fedde wrote:
>> On my sata attached ssd boot disk I see sequential write rates of:
>>
>> dd if=/dev/zero of=/tmp/foo bs=4k count=100000
>
> In the case above, the dd command is writing 4k blocks out to disk
> sequentially for 410 megabytes and it shows how one SSD disk compares to
> one SATAII disk.
>
> But change this benchmark to one where random blocks of that previously
> written 410MB file are changed in small amounts in the way a database
> server or a RAID array might have to make changes and the performance
> results should be drastically different.
>
> While a traditional disk can simply go to the 4k sector on disc and
> rewrite it, many SSDs have to read in a 'line' of 32k-64k, manipulate it
> in memory to accomodate the 4k change, and then re-write the 32k line
> back out to the disk.
>
> This is why you aren't expected to see RAID array vendors dropping disks
> for SSDs.  SSDs and SATA drives are apples and oranges, not better
> oranges.
>
> ps. Nice reading your results Chris, I'm always a fan of real-world,
> user-produced benchmarks.
>
> --
> Maxwell Spangler
>
> _______________________________________________
> clue-talk mailing list
> clue-talk at cluedenver.org
> http://www.cluedenver.org/mailman/listinfo/clue-talk
>



-- 
Nick Verbeck - NerdyNick
----------------------------------------------------
NerdyNick.com
SkeletalDesign.com
VivaLaOpenSource.com
Coloco.ubuntu-rocks.org