RAID 10 vs. RAID 5 question

I have an external RAID with 10 36GB 15K drives. These drives are only for
the data of the SQL server, not tran logs, tempdb's etc.
Database Type is OLTP with more then 30% writes then reads
This is my question:
I understand more spindles are better, and I understand RAID 10 is faster
for writes then a RAID5 configuration. Plus both configurations will give me
enough HD space. So on with the question.
Should i use RAID 10 or RAID5?
RAID 10 would give be theoretically faster writes and great protection, but
gives me only 5 disks to write to simultaneously.
RAID 5 gives me around 9 disks but is known to be slower due to the
overhead.
Is the RAID5 really that much slower that it would hinder performance
compared to a RAID 10 with only 5 disks?
So what do you guys think?
-King
yes.
go RAID 1+0
If you have time you can test this with a handful of IOStress test
utilities.
you could also talk directly to your vendor.
As I learned today, you will also want to maximize the WRITE Cache on the
RAID Controller(s).
Cheers
Greg Jackson
PDX, Oregon
|||Hi
Where are the transaction logs going? They are as critical to the Db as the
data files.
RAID-10 wins hands down. RAID-5 has just too much overhead for the parity.
From a safety perspective, on RAID-5, you loose 2 drives, bye bye data.
RAID -10 can lose half it's drives, as long as it is never both pairs of a
mirror.
So performance wins.
Regards
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland
IM: mike@.epprecht.net
MVP Program: http://www.microsoft.com/mvp
Blog: http://www.msmvps.com/epprecht/
"news.microsoft.com" <richk@.bluestreammedia.com> wrote in message
news:OUGbB56EFHA.2176@.TK2MSFTNGP15.phx.gbl...
> I have an external RAID with 10 36GB 15K drives. These drives are only
for
> the data of the SQL server, not tran logs, tempdb's etc.
> Database Type is OLTP with more then 30% writes then reads
> This is my question:
> I understand more spindles are better, and I understand RAID 10 is faster
> for writes then a RAID5 configuration. Plus both configurations will give
me
> enough HD space. So on with the question.
> Should i use RAID 10 or RAID5?
> RAID 10 would give be theoretically faster writes and great protection,
but
> gives me only 5 disks to write to simultaneously.
> RAID 5 gives me around 9 disks but is known to be slower due to the
> overhead.
> Is the RAID5 really that much slower that it would hinder performance
> compared to a RAID 10 with only 5 disks?
> So what do you guys think?
>
> -King
>
|||Hi
Yes, have write cache, as long as the RAID card has battery backup. If not,
kiss your data goodbye as you will have some decent corruption.
For us, data security over performance, so write caching is off.
Regards
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland
IM: mike@.epprecht.net
MVP Program: http://www.microsoft.com/mvp
Blog: http://www.msmvps.com/epprecht/
"pdxJaxon" <GregoryAJackson@.Hotmail.com> wrote in message
news:#62vk86EFHA.1264@.TK2MSFTNGP12.phx.gbl...
> yes.
> go RAID 1+0
> If you have time you can test this with a handful of IOStress test
> utilities.
> you could also talk directly to your vendor.
> As I learned today, you will also want to maximize the WRITE Cache on the
> RAID Controller(s).
>
> Cheers
> Greg Jackson
> PDX, Oregon
>
|||yes....HEAVENS YES.
One needs "battery backed" Cache.
GAJ
|||"Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
news:%23qs9yC7EFHA.2540@.TK2MSFTNGP09.phx.gbl...
> Hi
> Yes, have write cache, as long as the RAID card has battery backup. If
not,
> kiss your data goodbye as you will have some decent corruption.
>
Yes we have a 72-hour backup on all RAID controllers
[vbcol=seagreen]
> For us, data security over performance, so write caching is off.
> Regards
> --
> Mike Epprecht, Microsoft SQL Server MVP
> Zurich, Switzerland
> IM: mike@.epprecht.net
> MVP Program: http://www.microsoft.com/mvp
> Blog: http://www.msmvps.com/epprecht/
> "pdxJaxon" <GregoryAJackson@.Hotmail.com> wrote in message
> news:#62vk86EFHA.1264@.TK2MSFTNGP12.phx.gbl...
the
>
|||"Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
news:e4EdmB7EFHA.4004@.tk2msftngp13.phx.gbl...
> Hi
> Where are the transaction logs going? They are as critical to the Db as
the
> data files.
> RAID-10 wins hands down. RAID-5 has just too much overhead for the parity.
> From a safety perspective, on RAID-5, you loose 2 drives, bye bye data.
> RAID -10 can lose half it's drives, as long as it is never both pairs of a
> mirror.
> So performance wins.
Mike,
Thanks! Thats what i thought.. but what puzzled me was the spindols. Since
the RAID5 configuration would have more drives to spread the write acrossed
then the RAID 1+0 i wasn't sure if it mattered. Keep in mind i haven't got
the hardware yet... so no testing has been done. and yes i will do testing.
thanks for your input.
-King
[vbcol=seagreen]
> Regards
> --
> Mike Epprecht, Microsoft SQL Server MVP
> Zurich, Switzerland
> IM: mike@.epprecht.net
> MVP Program: http://www.microsoft.com/mvp
> Blog: http://www.msmvps.com/epprecht/
> "news.microsoft.com" <richk@.bluestreammedia.com> wrote in message
> news:OUGbB56EFHA.2176@.TK2MSFTNGP15.phx.gbl...
> for
faster[vbcol=seagreen]
give
> me
> but
>
|||In a Raid 1+0 you can't look at it as only x many drives to write to. In
your case you 10 drives that are configured as such. 5 Mirrored pairs that
are striped in a Raid 0 configuration. Yes that means you have to split the
data 5 ways vs. 9 for the Raid 5 but each split goes to a mirrored pair. The
mirrored pair has the option to read from one disk and write to the other,
write to both, read from both etc. It can be smart in how it reads and
writes to the mirrored pair. That plus the fact it doe not have to
calculate parity is a fast combination.
Andrew J. Kelly SQL MVP
"news.microsoft.com" <richk@.bluestreammedia.com> wrote in message
news:%23qHC657EFHA.3672@.TK2MSFTNGP14.phx.gbl...
> "Mike Epprecht (SQL MVP)" <mike@.epprecht.net> wrote in message
> news:e4EdmB7EFHA.4004@.tk2msftngp13.phx.gbl...
> the
> Mike,
> Thanks! Thats what i thought.. but what puzzled me was the spindols.
> Since
> the RAID5 configuration would have more drives to spread the write
> acrossed
> then the RAID 1+0 i wasn't sure if it mattered. Keep in mind i haven't
> got
> the hardware yet... so no testing has been done. and yes i will do
> testing.
> thanks for your input.
> -King
>
> faster
> give
>
|||This website has great details and arguments why you should not use RAID 5
for a RDBMS implementation
http://www.baarf.com/
GertD@.SQLDev.Net
Please reply only to the newsgroups.
This posting is provided "AS IS" with no warranties, and confers no rights.
You assume all risk for your use.
Copyright SQLDev.Net 1991-2005 All rights reserved.
"news.microsoft.com" <richk@.bluestreammedia.com> wrote in message
news:OUGbB56EFHA.2176@.TK2MSFTNGP15.phx.gbl...
>I have an external RAID with 10 36GB 15K drives. These drives are only for
> the data of the SQL server, not tran logs, tempdb's etc.
> Database Type is OLTP with more then 30% writes then reads
> This is my question:
> I understand more spindles are better, and I understand RAID 10 is faster
> for writes then a RAID5 configuration. Plus both configurations will give
> me
> enough HD space. So on with the question.
> Should i use RAID 10 or RAID5?
> RAID 10 would give be theoretically faster writes and great protection,
> but
> gives me only 5 disks to write to simultaneously.
> RAID 5 gives me around 9 disks but is known to be slower due to the
> overhead.
> Is the RAID5 really that much slower that it would hinder performance
> compared to a RAID 10 with only 5 disks?
> So what do you guys think?
>
> -King
>