| 1 |
Am 20.12.2009 08:14, schrieb Alan McKinnon: |
| 2 |
> On Saturday 19 December 2009 12:19:05 Stefan G. Weichinger wrote: |
| 3 |
>> I expected more WOW in terms of overall speed ... |
| 4 |
> |
| 5 |
> SSDs are not a magic bullet, it's unlikely they will give you a |
| 6 |
> killer performance improvement that makes you go "WOW!!!" |
| 7 |
|
| 8 |
sad to hear ;-) |
| 9 |
|
| 10 |
> SSDs suck at random writes. Typical usage scenario on a workstation |
| 11 |
> is lots of random writes compared to relatively few random reads - |
| 12 |
> reads tend not to be all that random as you re-read the same thing |
| 13 |
> often and it gets cached. |
| 14 |
> |
| 15 |
> Intel SSDs are far superior at random writes than any other SSD out |
| 16 |
> there but it's still nowhere near as optimised as spinning drives, |
| 17 |
> and kernels by and large are still optimised for spinning drives |
| 18 |
> too. |
| 19 |
|
| 20 |
Acknowledged. Just for reference, I switched over to the |
| 21 |
noop-IO-scheduler and checked that /sys/block/sdX/queue/rotational is |
| 22 |
detected/set correctly at value 0 for the ssd. These are two ssd-related |
| 23 |
modifications I know of ... |
| 24 |
|
| 25 |
> This may account for your overall feeling of under-whelmedness why |
| 26 |
> still seeing a significant boot-time speed up. You also have enough |
| 27 |
> RAM so that almost an entire typical workstation session could fit in |
| 28 |
> RAM and seldom touch the disk especially with a large interval |
| 29 |
> between disk syncs |
| 30 |
|
| 31 |
Yes, you are right. Thanks. |
| 32 |
|
| 33 |
What I still wonder: |
| 34 |
|
| 35 |
* I copied my gentoo over while the ssd still was on firmware 02HA, I |
| 36 |
upgraded to 02HD (latest Intel-firmware for the X25-M G2) with the os |
| 37 |
already on the drive. Does that somehow make a difference to data |
| 38 |
written *after* the upgrade? I just wonder if I should somehow restart |
| 39 |
from scratch with the new firmware, just to "do it right" or "get the |
| 40 |
optimum". |
| 41 |
|
| 42 |
* The new firmware supports TRIM. As far as I understand, the |
| 43 |
block-layer in the linux kernel 2.6.32 does not yet support it, |
| 44 |
correct? Any way to use that command with my current 2.6.32? ext4 |
| 45 |
supports it (I use ext4 for the root-partition) but I am not sure if |
| 46 |
that is enough when libata does not? |
| 47 |
|
| 48 |
When I google I get different results, some saying the kernel supports |
| 49 |
TRIM since 2.6.28 ... |
| 50 |
|
| 51 |
Hmm, good info here: |
| 52 |
|
| 53 |
http://www.linux-mag.com/id/7272 |
| 54 |
|
| 55 |
says: |
| 56 |
|
| 57 |
> So at the moment, ext4 informs the block layer that blocks that |
| 58 |
> belong to deleted files can be discard, so once TRIM-capable SSD’s |
| 59 |
> become available, and the Linux block layer actually sends the TRIM |
| 60 |
> command to the hard drives, everything will be all set to go. |
| 61 |
|
| 62 |
Does anyone have more recent info? |
| 63 |
|
| 64 |
Thanks, best greetings, have some peaceful days ... |
| 65 |
|
| 66 |
Stefan |
Archives