| 1 |
On 4/5/20 3:50 pm, hitachi303 wrote: |
| 2 |
> Am 04.05.2020 um 02:46 schrieb Rich Freeman: |
| 3 |
>> On Sun, May 3, 2020 at 6:50 PM hitachi303 |
| 4 |
>> <gentoo-user@××××××××××××××××.de> wrote: |
| 5 |
|
| 6 |
>> ... |
| 7 |
> So you are right. This is the way they do it. I used the term raid to |
| 8 |
> broadly. |
| 9 |
> But still they have problems with limitations. Size of room, what air |
| 10 |
> conditioning can handle and stuff like this. |
| 11 |
> |
| 12 |
> Anyway I only wanted to point out that there are different approaches |
| 13 |
> in the industries and saving the data at any price is not always |
| 14 |
> necessary. |
| 15 |
> |
| 16 |
I would suggest that once you go past a few drives there are better ways. |
| 17 |
|
| 18 |
I had two 4 disk, bcache fronted raid 10's on two PC'cs with critical |
| 19 |
data backed up between them. When an ssd bcache failed in one, and two |
| 20 |
backing stores in the other almost simultaneously I nearly had to resort |
| 21 |
to offline backups to restore the data ... downtime was still a major pain. |
| 22 |
|
| 23 |
I now have a 5 x cheap arm systems and a small x86 master with 7 disks |
| 24 |
across them - response time is good, power use seems less (being much |
| 25 |
cooler, quieter) than running two over the top older PC's. The |
| 26 |
reliability/recovery time (at least when I tested by manually failing |
| 27 |
drives and causing power outages) is much better. |
| 28 |
|
| 29 |
I am using moosefs, but lizardfs looks similar and both can offer |
| 30 |
erasure coding which gives more storage space still with recovery if you |
| 31 |
have enough disks. |
| 32 |
|
| 33 |
Downside is maintaining more systems, more complex networking and the |
| 34 |
like - Its been a few months now, and I wont be going back to raid for |
| 35 |
my main storage. |
| 36 |
|
| 37 |
BillK |
Archives