[K12OSN] Linux "Software RAID"

[Rob Owens]

Terrell Prude' Jr. wrote:
> Rob Owens wrote:
>> On Thu, Aug 07, 2008 at 03:20:21PM -0400, Terrell Prude' Jr. wrote:
>>  
>>> Carl Keil wrote:
>>>    
>>>> Hey Folks,
>>>>
>>>> I hear people extolling the virtues of "software RAID" on the list a 
>>>> lot.  I'm finally setting up a production server in a school and I 
>>>> have enough disks to play with to do RAID.  I'm leaning towards RAID 
>>>> 5.  Anyway, when people say "software RAID" do they mean just 
>>>> setting up a RAID in LVM Manager?  Or is the mdadm command the 
>>>> simpler, more robust, preferred way to do this?  I never thought 
>>>> about using LVM for this before, but the last time I was in there I 
>>>> noticed some RAID options.  This is for a Samba/LDAP/home directory 
>>>> server.
>>>>
>>>> Thanks,
>>>>
>>>> ck
>>>>       
>>> "Software RAID" is simply doing the RAID (striping, mirroring, 
>>> parity, whatever) in the OS instead of on a dedicated card.  Windows 
>>> NT, from at least v3.50 (way back when), can do this, and Linux can 
>>> do it as well.
>>>
>>> If you're thinking of RAID 5, which is my preferred level, I'd avoid 
>>> doing it in software and instead opt for a dedicated RAID card.  
>>> Something like an LSI MegaRAID 150-6 SATA controller.  If you do it 
>>> in software, you'll eat up some CPU doing the parity calculations, so 
>>> you definitely want to offload that.  However, for just mirroring 
>>> (say, RAID 1), you should be fine, because the CPU hit for mirroring 
>>> is minimal.
>>>
>>>     
>> I hear lots of people talk about the CPU hit of software RAID.  But 
>> how much hit is there really?  Suppose for argument's sake I can get a 
>> hardware RAID card for $100.  If I instead used software RAID and 
>> spent my $100 on a better CPU, wouldn't I be ahead of the game?
>>
>> -Rob
>>   
> 
> No, I don't believe so.  For one thing, as Dan Young put it, it's much 
> easier to deal with swapping a failed disk out with a dedicated card.  
> That by itself is a *BIG DEAL*.  Additionally, if you do have a disk 
> fail, your CPU will take an especially big hit, because then it's got to 
> reconstruct data from the parity info for *all* disk accesses, not just 
> writes.  Oops....
> 
I realize the ease-of-use advantage of hardware RAID, but I really was 
talking about the performance hit only.

Realistically, a hardware RAID costs about $300, and that could buy me a 
2nd 4-core processor.  I have to believe that a modern 4-core processor 
is way more capable of handling a RAID rebuild than whatever chip is 
onboard a hardware RAID controller.  Of course, I haven't done any 
testing...

Even though we've debated software vs. hardware RAID on this list many 
times, these discussions always hold my interest.

-Rob
********************************************************

The information transmitted is intended only for the person or entity to
which it is addressed and may contain confidential and/or privileged
material. If you are not the addressee, any disclosure, reproduction,
copying, distribution, or other dissemination or use of this transmission in
error please notify the sender immediately and then delete this e-mail.
E-mail transmission cannot be guaranteed to be secure or error free as
information could be intercepted, corrupted lost, destroyed, arrive late or
incomplete, or contain viruses.
The sender therefore does not accept liability for any errors or omissions
in the contents of this message which arise as a result of e-mail
transmission. If verification is required please request a hard copy
version.

********************************************************