SCSI drives recommendations

Anthony David adavid at adavid.com.au
Tue Jan 22 12:50:25 EST 2002


andrew at bishop.dropbear.id.au writes:

> On Tue, 22 Jan 2002, Steven Hanley wrote:
> 
> [...]
> > I dont know that the calculation of parity is a slow down, I notice that when
> > you insmod the raid5 module it does some tests with different code to see what
> > speed it can do the parity calculations at.
> 
> Theory is good, and in theory, there is no difference between theory and
> practice - but in practice, there is.

The major issue with RAID 5 is writing. Performance is generally abysmal
once your write/read ratio goes up. You are doing two writes instead of
one. The only way to counter that is to have non-volatile write cache
which is generally only practical in a hardware RAID device.


> -- attribution forgotten, sorry :)
> 
> I've played around with various raids on my (ide) system at home.  3 fast
> ide disks in there (each gets about 33MB/s read).  raid 0 (striping only,
> no redundancy) across 2 of them gave me 66MB/s - no slowdown at all.
> That's whay I'm currently using.
> 
> Now, those 2 disks are 60G each and empty.  The third disk is smaller,
> and also has my root partition on it, so I couldn't just try raid 5
> without losing lots of capacity.  Still, I wanted to play around with raid
> 5 and see how it performed, so I used a spare couple of G from the third
> disk, and set up a small raid 5.  I only got about 15 MB/s read from it -
> less than 1/4 of what it "should" have been.
> 
> As I have no real need for redundancy at home (and only 2 disks to use
> for it anyway), I just run with raid 0.
> 
> Andrew
> 
> 

-- 
Anthony David

Gambling(n): A discretionary tax on those asleep during high school maths
http://adavid.com.au/
0xA72CE1ED fingerprint = EA1E C69E FE59 BBE1 AA4B  F354 BD09 9765 A72C E1ED




More information about the linux mailing list