[mythtv-users] Fastest RAID for HD?

Markus Schulz msc at antzsystem.de
Sat Feb 16 15:38:33 UTC 2008


Am Samstag, 16. Februar 2008 schrieb Roger Heflin:
> Markus Schulz wrote:
> > Am Freitag, 15. Februar 2008 schrieb Marc:
> > [...]
> >
> >> Don't assume people on this list are experts. Some are, some are
> >> informed users who have done research, some just make assumptions
> >> based off of one post or one article they read.
> >>
> >> >From a statistical standpoint..
> >>
> >> Raid0 would be the fastest, it is common sense. Your speed for any
> >> action would be dependent on the number of drives you had or N.
> >> However if one drive fails you lose all data.
> >>
> >> Raid3 or 5 would be the second fastest and would be safer due to
> >> the ability to continue working without loss of data if you had a
> >> single drive fail. The difference between these 2 raid types is
> >> the way they setup parity. Raid3 has a dedicated parity drive.
> >> Raid5 spreads parity across all the drives.
> >> Speed of any action on either of these raid configs would be the
> >> number of drives you had minus 1 or N-1.
> >
> > sorry but thats not true. Show me a raid5 with 5-6 drives which
> > will got
> >
> >> 300MB/s _write_ performance...
> >
> > Fastest Raid _WITH_ redundancy was raid10.
>
> The high dollar hardware raid controllers will get close to the N-1
> rate, I have benchmarked them under sustained load, but the hardware
> raid are in the $500-$1000 range, and raid10 with 6 disks will in a
> perfect world be 180-210mb/second (3 effective
> drives-60-70mb/second/drive).   

i've built a system two weeks ago with:

6 * 73Gb 15k 2.5" sas in a smart array p400 (512MB cache) for a HP dl380 
G5 system

d1:/mnt/disk2# dd if=/dev/zero of=testfile bs=1M count=16384
16384+0 Datensätze ein
16384+0 Datensätze aus
17179869184 Bytes (17 GB) kopiert, 55,6692 Sekunden, 309 MB/s

d1:/mnt/disk2# dd if=testfile of=/dev/null bs=1M
16384+0 Datensätze ein
16384+0 Datensätze aus
17179869184 Bytes (17 GB) kopiert, 56,1641 Sekunden, 306 MB/s

-- 
Markus Schulz


More information about the mythtv-users mailing list