RAID Levels 5 vs. 0+1

RAID Levels 5 vs. 0+1

Post by Bruce Walto » Mon, 20 May 1996 04:00:00



We are planning on setting up a RAID on one of our Netware servers.  As
we want striping and data redundancy, we have limited our choice to RAID
levels 0+1 or 5.  The documentation that came with our controller
explains that 0+1 uses striping and mirroring while 5 uses striping and
parity.  It does not give any other differences.  If there are no other
differences, then RAID 5 wins hands down as we would be able to use 75%
of our drive storage for data as opposed to 50% for 0+1.

Does anyone know of any other trade offs?  Is 0+1 faster and or safer?

Bruce Walton
Info. Sys. Analyst
CA Conservation Corps

 
 
 

RAID Levels 5 vs. 0+1

Post by Neil Waybrig » Tue, 21 May 1996 04:00:00


0 + 1 is  faster, and it is safer.  Since there is 100% redundancy,
you can withstand more disk failures before you lose any data.  Since
there is no parity to calculate, nor the the parity location to
determine, the controller hardware/software is simpler (and usually
faster).  Simple striping is the fastest way to get data off a disk,
and 0 + 1 merely makes that safe.  It also introduces the possibility
of round-robin reads that can nearly double your I/O rates in a nearly
exclusively read situation. The old rule of thumb is:

Fast, safe, cheap.... Pick any two.....
Fast and safe is 0 + 1   (not cheap since it has a 100% disk overhead)
fast and cheap is RAID 0    (not safe since the loss of any drive
means the loss of all data)
cheap and safe is RAID 3 or 5   (not fast due to the requirement to
generate parity on both, the single tasking nature of RAID 3 or the
small write penalty of RAID 5)

Neil


>We are planning on setting up a RAID on one of our Netware servers.  As
>we want striping and data redundancy, we have limited our choice to RAID
>levels 0+1 or 5.  The documentation that came with our controller
>explains that 0+1 uses striping and mirroring while 5 uses striping and
>parity.  It does not give any other differences.  If there are no other
>differences, then RAID 5 wins hands down as we would be able to use 75%
>of our drive storage for data as opposed to 50% for 0+1.
>Does anyone know of any other trade offs?  Is 0+1 faster and or safer?
>Bruce Walton
>Info. Sys. Analyst
>CA Conservation Corps



 
 
 

RAID Levels 5 vs. 0+1

Post by sarf.. » Wed, 22 May 1996 04:00:00



> We are planning on setting up a RAID on one of our Netware servers.  As
> we want striping and data redundancy, we have limited our choice to RAID
> levels 0+1 or 5.  The documentation that came with our controller
> explains that 0+1 uses striping and mirroring while 5 uses striping and
> parity.  It does not give any other differences.  If there are no other
> differences, then RAID 5 wins hands down as we would be able to use 75%
> of our drive storage for data as opposed to 50% for 0+1.

> Does anyone know of any other trade offs?  Is 0+1 faster and or safer?

> Bruce Walton
> Info. Sys. Analyst
> CA Conservation Corps


The trade-off is speed vs. capacity: RAID 0, 1 or 0+1 are faster for both Read
and Write operations. Read operations can be split between the mirrored disks
and Write operations are simpler as well. In a RAID-5 system every Write
operation requires reading data from two disks and then writing to the same
two disks.

Another trade-off for Netware is that mirroring is supported directly by the
NOS while RAID-5 requires a smart controller (with on-board CPU and memory).

                                Regards
                                Assaf

 
 
 

RAID Levels 5 vs. 0+1

Post by Waldec » Wed, 22 May 1996 04:00:00


With RAID 5 you have an write penalty. For one update you have four I/O's.

1. Read Record
2. Read old parity
3. Write Record
4. Write new parity
How fast this works, depends on the vendor of the RAID 5 system.

Klaus Waldeck
StorageTek Germany

 
 
 

RAID Levels 5 vs. 0+1

Post by John Hasca » Wed, 22 May 1996 04:00:00



}With RAID 5 you have an write penalty. For one update you have four I/O's.
}1. Read Record
}2. Read old parity
}3. Write Record
}4. Write new parity

     If the write is large enough there is only 1 (as in N+1)
     extra write for the parity.  In any event, any decent
     RAID-5 will have a cache which will hopefully eliminate
     the 2 reads a significant portion of the time.

     In my experience (with Digital's RAID Array 200 product
     and Fast Wide 7200rpm disks) the RAID 5 config was quite
     competetive with RAID 0+1, and in any event was capable of
     moving data faster than the FDDI interface -- so, this being
     on a fileserver, there is little incentive to do RAID 0+1.

John
--
John Hascall                ``An ill-chosen word is the fool's messenger.''
Moderator, comp.unix.wizards
Systems Software Engineer, ISU Comp Center + Ames, IA  50011 + 515/294-9551
<a href="http://www.cc.iastate.edu/staff/systems/john/">My Homepage</a>

 
 
 

RAID Levels 5 vs. 0+1

Post by Dick Wilm » Wed, 22 May 1996 04:00:00



>The trade-off is speed vs. capacity: RAID 0, 1 or 0+1 are faster for both Read
>and Write operations. Read operations can be split between the mirrored disks
>and Write operations are simpler as well. In a RAID-5 system every Write
>operation requires reading data from two disks and then writing to the same
>two disks.

RAID 5 only needs to read old data or parity for short writes. For a long
write RAID 5 can write N units of data and newly computed parity in parallel.
For long reads RAID 5 is architecturally the highest bandwidth available since
it can read all data in parallel, skipping parity (using the same number of
drives in all cases).

--
                       * Wilmot
                        Diablo Data Design

 
 
 

1. Oracle Redo/Archive Logs w/T3 Raid 0+1 vs Raid 5

I know im not the first to ask this but can anyone here share there
recommendations ?  Sun is telling us to use Raid 5 for Oracle redo/archive
logs but our DBA's is pushing for 0+1 with partner pairs which consumes way
two much disks I dont think it justify the cost. Anyone have any
recommendations ? were using 6960 iirc 1gig of cache per brick.

- Rodrick Brown
- Unix SA

2. Only 16colors

3. More on RAID 0+1 ("RAID 10")

4. kosko's FAM

5. Expanding Raid 0+1

6. SMS Site Comp. manager Freaked Out!!

7. RAID 7 Is An Extension Of The 5 Berkeley RAID Levels

8. @MailSend in Lotus Script

9. High Level Assemblers vs. High Level Language Compilers

10. Theoretical Question...RAID 10 vs RAID 5

11. RAID-10 vs. RAID-01 ?

12. Hardware RAID vs. Software RAID

13. RAID 7 vs. Earlier Raid 4