[Klug-general] Software Vs Hardware Raid

[MacGyveR]

On Friday 04 Jul 2008, Paul Littlefield wrote:
> On Thursday 03 July 2008 16:24:07 J D Freeman wrote:
> > These days linux software raid is your best option.
>
> Yes, I have to agree. Used it for years now and it seems the best
> trade-off.
>
> There is a slight performance hit with RAID 1 (mirroring) but the
> piece-of-mind outways that.
>
> Here is some RAID output from one of my clients' servers. It has 5 drives
> which comprise: 2 in use, 2 hot spares and 1 cold spare in the case ready
> to go.
>
> server2 ~ # cat /proc/mdstat
> Personalities : [raid1]
> md1 : active raid1 sdd1[2](S) sdc1[3](S) sdb1[1] sda1[0]
>       24000 blocks [2/2] [UU]
>
> md2 : active raid1 sdd2[2](S) sdc2[3](S) sdb2[1] sda2[0]
>       987904 blocks [2/2] [UU]
>
> md3 : active raid1 sdd3[2](S) sdc3[3](S) sdb3[1] sda3[0]
>       48837504 blocks [2/2] [UU]
>
> md4 : active raid1 sdd4[2](S) sdc4[3](S) sdb4[1] sda4[0]
>       194346240 blocks [2/2] [UU]
>
>
>
> It also has SMART health monitoring on the SATA drives themselves:-
>
> server2 ~ # smartctl -d ata -H /dev/sda
> smartctl version 5.37 [x86_64-pc-linux-gnu] Copyright (C) 2002-6 Bruce
> Allen Home page is http://smartmontools.sourceforge.net/
>
> === START OF READ SMART DATA SECTION ===
> SMART overall-health self-assessment test result: PASSED
> Please note the following marginal Attributes:
> ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED 
> WHEN_FAILED RAW_VALUE 190 Temperature_Celsius     0x0022   067   045   045 
>   Old_age   Always   In_the_past 33
>
>
>
> Hope this helps.
>
> Regards
>
> PAULLY
>
> _______________________________________________
> Kent mailing list
> Kent at mailman.lug.org.uk
> https://mailman.lug.org.uk/mailman/listinfo/kent


here is my thoughts, in no real order on this subject

raid is all about high availability you have multiple disks so if one dies 
then others continue to work. it not always the disk that dies. back in the 
day when pata drives were the norm a m/b controller could die and take out 
both disks on the channel. this is why in software raid you should only use 1 
device per channel eg hda & hdc. luckly with sata this problem has gone away. 

if you are using hardware or software raid and your controller dies then you 
can just buy a new one, but with hardware raid you will need to get hold of 
the same controller. if you are using any type of interface card such as 
sata/pata/scsi a backup should be purchased or you are just moving the 
problem somewhere else. m/b these days come with 2 or 4 sata/pata ports, if 
this is not enough you will have to buy additional cards from different 
brands if possible, sata cards are cheap at about £15 so there is no reason 
not to buy a spare, but saying that would you trust your data on a £15 sata 
card? if you are using scsi, having all your disks run from the same 
controller, is not optimal but is all about doing the best you can with the £ 
available.

software raid is slower by its nature, but most servers will not notice a 
difference. if you are concerned about the speed then there are lots of other 
settings to tune to get the most out of any setup.

when buying drives, do not buy all the drives from the same vendor as a 
production fault could take out the entire array.

-- 
--------------------------------
http://www.thedumbterminal.co.uk