RAID0 vs. RAID5 - Benchmark

about how many percent goes the benchmarkable speed down when I change a
RAID0 Sytem with 4 hard disks to a RAID5 System?

Ingo

Ingo Seibold wrote:

about how many percent goes the benchmarkable speed down when I change a
RAID0 Sytem with 4 hard disks to a RAID5 System?

Ingo

Depends a lot on the controller. RAID5 will be slower to write unless you
have an extremely evil controller, but with a good RAID5 controller, reads
could match those of a RAID0 system, although the calculation would be:

RAID0 speed with X disks == RAID5 speed with X+1 disks.

At least that's what I'm told

Either way, the security provided by RAID5 is invaluable, the SX4000 has
been my best spent money in a long time!

--
I win!

Previously Ingo Seibold wrote:
about how many percent goes the benchmarkable speed down when I change a
RAID0 Sytem with 4 hard disks to a RAID5 System?

If you have reasonably fast disks the read speed will be limited by
the PCI speed, unless you have 66MHz PCI, 64 bit PCI or PCI-E and a
matching RAID controller. Write speed depends on the card. I have had
an Adaptec SATA RAID5 card reading and writing slower than Linux
software RAID.

In short: Reading should be about the same, maybe a bit (25%) slower.
Writing is impossible to tell without knoing the details of the
RAID% and the RAID0 solution being compared. Usually knowing
the details will still not be enough and actual benchmarking
is the only option.

And of course it depends on the specific benchmark used.

Arno
--
For email address: lastname AT tik DOT ee DOT ethz DOT ch
GnuPG: ID:1E25338F FP:0C30 5782 9D93 F785 E79C 0296 797F 6B50 1E25 338F
"The more corrupt the state, the more numerous the laws" - Tacitus

Arno Wagner wrote:

Previously Ingo Seibold wrote:
about how many percent goes the benchmarkable speed down when I change a
RAID0 Sytem with 4 hard disks to a RAID5 System?

If you have reasonably fast disks the read speed will be limited by
the PCI speed, unless you have 66MHz PCI, 64 bit PCI or PCI-E and a
matching RAID controller.

Depends on the size of the array and the data pattern. If the data is such
that most reads are sequential then the bus could be the limiting factor.
If most reads are of random sectors then access time is still going to
dominate.

Write speed depends on the card. I have had
an Adaptec SATA RAID5 card reading and writing slower than Linux
software RAID.

In short: Reading should be about the same, maybe a bit (25%) slower.
Writing is impossible to tell without knoing the details of the
RAID% and the RAID0 solution being compared. Usually knowing
the details will still not be enough and actual benchmarking
is the only option.

And of course it depends on the specific benchmark used.

Arno

--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)