IDEA: Shortening boot-time

[Russell Coker]

On Mon, 26 Jul 2004 18:05, Mark Lane <mark at harddata.com> wrote:
> On July 24, 2004 05:03 am, Russell Coker <russell at coker.com.au> wrote:
> > Last time I did some tests on Linux software RAID I found that it was
> > lacking in this regard.  I would have hoped to see some read benchmarks
> > showing that a RAID-1 with two disks is nearly twice as fast as a single
> > disk, however I didn't find any test that showed such a result or
>
> You actually expect to get twice the performance out of a RAID 1 array in
> software RAID? First of all writes are not going to be any faster than a

On some tests, yes.  When I have a file system on a RAID-1 with two large 
files (significantly larger than ram) then I should (theoretically) be able 
to read them both (with one process reading each file) in the same time that 
a non-RAID system could read one file with one process.  In practice I would 
like to see 90% of that performance.

The fact that such results seemed impossible to achieve indicate a scheduling 
problem with Linux software RAID and possibly block devices in general.

> single drive and could be slower depending on your hardware. Reads could be
> faster depending on you hardware but twice? Only very highend (expensive)
> SCSI raid cards do that and only theoretically. Practically you won't get
> twice the performance.

What's so special about SCSI RAID cards?  They have a CPU (which is much 
slower than recent Intel or Athlon CPUs), some RAM, and an embedded OS that 
manages RAID, caching, etc.  Anything that such cards can do Linux should 
also be able to do.

It wouldn't surprise me if some of those cards RAN Linux.

-- 
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page