Folks, I am setting up a new server and was wondering of the recommended
drive setup. I have done the following:

2 36GB Drives - OS
2 72GB Drives - T-Logs
4 72GB Drives - DB / Backups

Everything is RAID 1+0 except the DB drives. They are RAID 5. However, I am
wondering if they too should be RAID 1+0. There is always this argument
about it versus RAID 5.

The databases are not big at all. This is a Departmental DB Server. I have
approximately 10GB of data and 100MB of Logs. Not large at all.

Please advise.

Thanks.
-wp

Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is more
fault-tolerant than RAID 5 and performs better on writes.

BTW, your OS and log drives are technically RAID 1 rather than 0+1 because
no striping is involved.

--
Hope this s.

Dan Guzman
SQL Server MVP

"news.microsoft.com" <not@thistime.com> wrote in message
news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl...
> Folks, I am setting up a new server and was wondering of the recommended
> drive setup. I have done the following:
>
> 2 36GB Drives - OS
> 2 72GB Drives - T-Logs
> 4 72GB Drives - DB / Backups
>
> Everything is RAID 1+0 except the DB drives. They are RAID 5. However, I
> am
> wondering if they too should be RAID 1+0. There is always this argument
> about it versus RAID 5.
>
> The databases are not big at all. This is a Departmental DB Server. I have
> approximately 10GB of data and 100MB of Logs. Not large at all.
>
> Please advise.
>
> Thanks.
> -wp
>
>

Hi Dan,

I'm not a SQL expert by any stretch of the imagination, but I thought
that it was advised to have MDF files on a RAID-1 or RAID-10 array,
and LDF files (including tempdb) on a RAID-10 for optimal performance,
with emphasis being on the LDFs residing on a RAID-10, as they are
more I/O intensive.

I work for a managed hosting company with our own staff of DBAs, so
I'm not usually the one making these kind of recommendations.

Thanks,
JL!

On Feb 11, 9:47 am, "Dan Guzman" <guzma...@nospam-
online.sbcglobal.net> wrote:
> Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is more
> fault-tolerant than RAID 5 and performs better on writes.
>
> BTW, your OS and log drives are technically RAID 1 rather than 0+1 because
> no striping is involved.
>
> --
> Hope this s.
>
> Dan Guzman
> SQL Server MVP
>
> "news.microsoft.com" <n...@thistime.com> wrote in message
>
> news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl...
>
>
>
> > Folks, I am setting up a new server and was wondering of the recommended
> > drive setup. I have done the following:
>
> > 2 36GB Drives - OS
> > 2 72GB Drives - T-Logs
> > 4 72GB Drives - DB / Backups
>
> > Everything is RAID 1+0 except the DB drives. They are RAID 5. However, I
> > am
> > wondering if they too should be RAID 1+0. There is always this argument
> > about it versus RAID 5.
>
> > The databases are not big at all. This is a Departmental DB Server. I have
> > approximately 10GB of data and 100MB of Logs. Not large at all.
>
> > Please advise.
>
> > Thanks.
> > -wp- Hide quoted text -
>
> - Show quoted text -

Ultimately, it comes down the I/O requirements of the log writes. Your log
I/O requirements may be such that RAID1 can provide sufficient performance.
It's not very meaningful to consider whether to place the log file on RAID1
or RAID10 in abstract.

But then, in reality decisions such as this often have to be made without
any knowledge of the real I/O requirements anyway. In that case, other
factors may figure large. One such factor is the size of your log file. Note
that if you configure RAID10 for the log file in this case, you would end up
with at least 144GB for the RAID10 drive, which may be too large for your log
file and leaves too little space for the data files.

Linchi

"Jeremy Lyons" wrote:

> Hi Dan,
>
> I'm not a SQL expert by any stretch of the imagination, but I thought
> that it was advised to have MDF files on a RAID-1 or RAID-10 array,
> and LDF files (including tempdb) on a RAID-10 for optimal performance,
> with emphasis being on the LDFs residing on a RAID-10, as they are
> more I/O intensive.
>
> I work for a managed hosting company with our own staff of DBAs, so
> I'm not usually the one making these kind of recommendations.
>
> Thanks,
> JL!
>
> On Feb 11, 9:47 am, "Dan Guzman" <guzma...@nospam-
> online.sbcglobal.net> wrote:
> > Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is more
> > fault-tolerant than RAID 5 and performs better on writes.
> >
> > BTW, your OS and log drives are technically RAID 1 rather than 0+1 because
> > no striping is involved.
> >
> > --
> > Hope this s.
> >
> > Dan Guzman
> > SQL Server MVP
> >
> > "news.microsoft.com" <n...@thistime.com> wrote in message
> >
> > news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl...
> >
> >
> >
> > > Folks, I am setting up a new server and was wondering of the recommended
> > > drive setup. I have done the following:
> >
> > > 2 36GB Drives - OS
> > > 2 72GB Drives - T-Logs
> > > 4 72GB Drives - DB / Backups
> >
> > > Everything is RAID 1+0 except the DB drives. They are RAID 5. However, I
> > > am
> > > wondering if they too should be RAID 1+0. There is always this argument
> > > about it versus RAID 5.
> >
> > > The databases are not big at all. This is a Departmental DB Server. I have
> > > approximately 10GB of data and 100MB of Logs. Not large at all.
> >
> > > Please advise.
> >
> > > Thanks.
> > > -wp- Hide quoted text -
> >
> > - Show quoted text -
>
>
>

As Linchi mentioned, RAID-1 is most commonly used for logs mostly because
log files aren't large enough to warrant multiple disks. Furthermore,
striping log files using RAID-10 probably won't provide a performance
benefit except in the most extreme write-intensive apps. Data files are
more likely to benefit from the extra spindles and striping you get with
RAID 10.

--
Hope this s.

Dan Guzman
SQL Server MVP

"Jeremy Lyons" <i.am.jammo@gmail.com> wrote in message
news:1171219025.439739.10700@v45g2000cwv.googlegro ups.com...
> Hi Dan,
>
> I'm not a SQL expert by any stretch of the imagination, but I thought
> that it was advised to have MDF files on a RAID-1 or RAID-10 array,
> and LDF files (including tempdb) on a RAID-10 for optimal performance,
> with emphasis being on the LDFs residing on a RAID-10, as they are
> more I/O intensive.
>
> I work for a managed hosting company with our own staff of DBAs, so
> I'm not usually the one making these kind of recommendations.
>
> Thanks,
> JL!
>
> On Feb 11, 9:47 am, "Dan Guzman" <guzma...@nospam-
> online.sbcglobal.net> wrote:
>> Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is more
>> fault-tolerant than RAID 5 and performs better on writes.
>>
>> BTW, your OS and log drives are technically RAID 1 rather than 0+1
>> because
>> no striping is involved.
>>
>> --
>> Hope this s.
>>
>> Dan Guzman
>> SQL Server MVP
>>
>> "news.microsoft.com" <n...@thistime.com> wrote in message
>>
>> news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl...
>>
>>
>>
>> > Folks, I am setting up a new server and was wondering of the
>> > recommended
>> > drive setup. I have done the following:
>>
>> > 2 36GB Drives - OS
>> > 2 72GB Drives - T-Logs
>> > 4 72GB Drives - DB / Backups
>>
>> > Everything is RAID 1+0 except the DB drives. They are RAID 5. However,
>> > I
>> > am
>> > wondering if they too should be RAID 1+0. There is always this argument
>> > about it versus RAID 5.
>>
>> > The databases are not big at all. This is a Departmental DB Server. I
>> > have
>> > approximately 10GB of data and 100MB of Logs. Not large at all.
>>
>> > Please advise.
>>
>> > Thanks.
>> > -wp- Hide quoted text -
>>
>> - Show quoted text -
>
>

@both-

I understand that you would not arbitrarily pick a RAID setup. My
misunderstanding comes in that I thought the log files were where the
high I/O occurs, and that your MDF files can be put on a lower
performance array because transactions aren't committed until (using
the full recovery model) a t-log backup occurs or a checkpoint is
created.

Is that correct?

On Feb 11, 1:22 pm, "Dan Guzman" <guzma...@nospam-
online.sbcglobal.net> wrote:
> As Linchi mentioned, RAID-1 is most commonly used for logs mostly because
> log files aren't large enough to warrant multiple disks. Furthermore,
> striping log files using RAID-10 probably won't provide a performance
> benefit except in the most extreme write-intensive apps. Data files are
> more likely to benefit from the extra spindles and striping you get with
> RAID 10.
>
> --
> Hope this s.
>
> Dan Guzman
> SQL Server MVP
>
> "Jeremy Lyons" <i.am.ja...@gmail.com> wrote in message
>
> news:1171219025.439739.10700@v45g2000cwv.googlegro ups.com...
>
>
>
> > Hi Dan,
>
> > I'm not a SQL expert by any stretch of the imagination, but I thought
> > that it was advised to have MDF files on a RAID-1 or RAID-10 array,
> > and LDF files (including tempdb) on a RAID-10 for optimal performance,
> > with emphasis being on the LDFs residing on a RAID-10, as they are
> > more I/O intensive.
>
> > I work for a managed hosting company with our own staff of DBAs, so
> > I'm not usually the one making these kind of recommendations.
>
> > Thanks,
> > JL!
>
> > On Feb 11, 9:47 am, "Dan Guzman" <guzma...@nospam-
> > online.sbcglobal.net> wrote:
> >> Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is more
> >> fault-tolerant than RAID 5 and performs better on writes.
>
> >> BTW, your OS and log drives are technically RAID 1 rather than 0+1
> >> because
> >> no striping is involved.
>
> >> --
> >> Hope this s.
>
> >> Dan Guzman
> >> SQL Server MVP
>
> >> "news.microsoft.com" <n...@thistime.com> wrote in message
>
> >>news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl.. .
>
> >> > Folks, I am setting up a new server and was wondering of the
> >> > recommended
> >> > drive setup. I have done the following:
>
> >> > 2 36GB Drives - OS
> >> > 2 72GB Drives - T-Logs
> >> > 4 72GB Drives - DB / Backups
>
> >> > Everything is RAID 1+0 except the DB drives. They are RAID 5. However,
> >> > I
> >> > am
> >> > wondering if they too should be RAID 1+0. There is always this argument
> >> > about it versus RAID 5.
>
> >> > The databases are not big at all. This is a Departmental DB Server. I
> >> > have
> >> > approximately 10GB of data and 100MB of Logs. Not large at all.
>
> >> > Please advise.
>
> >> > Thanks.
> >> > -wp- Hide quoted text -
>
> >> - Show quoted text -- Hide quoted text -
>
> - Show quoted text -

Log file I/O is mostly sequential writes. Log performance is important
mostly because applications must wait for the log I/O to actually complete
when transactions are committed (implicit or explicit COMMIT). Separating
data and log files optimizes log writes, improving both response time and
throughput.

SQL Server uses write-ahead logging so data are always written to the log
before data files. The data file writes happen asynchronously (lazywriter,
checkpoint) and can occur either before or after the transaction is actually
committed. You are right that you could place data files on
lower-performance array as long as your application isn't write-intensive.
However, data files are more likely to benefit from striping than log files
because sequential scan performance is improved and distributing the
workload among more drives improves random read and write performance.

--
Hope this s.

Dan Guzman
SQL Server MVP

"Jeremy Lyons" <i.am.jammo@gmail.com> wrote in message
news:1171228008.486359.40280@v45g2000cwv.googlegro ups.com...
> @both-
>
> I understand that you would not arbitrarily pick a RAID setup. My
> misunderstanding comes in that I thought the log files were where the
> high I/O occurs, and that your MDF files can be put on a lower
> performance array because transactions aren't committed until (using
> the full recovery model) a t-log backup occurs or a checkpoint is
> created.
>
> Is that correct?
>
> On Feb 11, 1:22 pm, "Dan Guzman" <guzma...@nospam-
> online.sbcglobal.net> wrote:
>> As Linchi mentioned, RAID-1 is most commonly used for logs mostly because
>> log files aren't large enough to warrant multiple disks. Furthermore,
>> striping log files using RAID-10 probably won't provide a performance
>> benefit except in the most extreme write-intensive apps. Data files are
>> more likely to benefit from the extra spindles and striping you get with
>> RAID 10.
>>
>> --
>> Hope this s.
>>
>> Dan Guzman
>> SQL Server MVP
>>
>> "Jeremy Lyons" <i.am.ja...@gmail.com> wrote in message
>>
>> news:1171219025.439739.10700@v45g2000cwv.googlegro ups.com...
>>
>>
>>
>> > Hi Dan,
>>
>> > I'm not a SQL expert by any stretch of the imagination, but I thought
>> > that it was advised to have MDF files on a RAID-1 or RAID-10 array,
>> > and LDF files (including tempdb) on a RAID-10 for optimal performance,
>> > with emphasis being on the LDFs residing on a RAID-10, as they are
>> > more I/O intensive.
>>
>> > I work for a managed hosting company with our own staff of DBAs, so
>> > I'm not usually the one making these kind of recommendations.
>>
>> > Thanks,
>> > JL!
>>
>> > On Feb 11, 9:47 am, "Dan Guzman" <guzma...@nospam-
>> > online.sbcglobal.net> wrote:
>> >> Consider RAID 10 for your data drives instead of RAID 5. RAID 10 is
>> >> more
>> >> fault-tolerant than RAID 5 and performs better on writes.
>>
>> >> BTW, your OS and log drives are technically RAID 1 rather than 0+1
>> >> because
>> >> no striping is involved.
>>
>> >> --
>> >> Hope this s.
>>
>> >> Dan Guzman
>> >> SQL Server MVP
>>
>> >> "news.microsoft.com" <n...@thistime.com> wrote in message
>>
>> >>news:%23xG4c6eTHHA.4744@TK2MSFTNGP02.phx.gbl.. .
>>
>> >> > Folks, I am setting up a new server and was wondering of the
>> >> > recommended
>> >> > drive setup. I have done the following:
>>
>> >> > 2 36GB Drives - OS
>> >> > 2 72GB Drives - T-Logs
>> >> > 4 72GB Drives - DB / Backups
>>
>> >> > Everything is RAID 1+0 except the DB drives. They are RAID 5.
>> >> > However,
>> >> > I
>> >> > am
>> >> > wondering if they too should be RAID 1+0. There is always this
>> >> > argument
>> >> > about it versus RAID 5.
>>
>> >> > The databases are not big at all. This is a Departmental DB Server.
>> >> > I
>> >> > have
>> >> > approximately 10GB of data and 100MB of Logs. Not large at all.
>>
>> >> > Please advise.
>>
>> >> > Thanks.
>> >> > -wp- Hide quoted text -
>>
>> >> - Show quoted text -- Hide quoted text -
>>
>> - Show quoted text -
>
>

I'd definitely go with RAID 10 for the database files. I'd keep the
backups on a separate drive from the databases. Not good if your backups
go with the databases when the drive fails...

Regards,
Hank Arnold

news.microsoft.com wrote:
> Folks, I am setting up a new server and was wondering of the recommended
> drive setup. I have done the following:
>
> 2 36GB Drives - OS
> 2 72GB Drives - T-Logs
> 4 72GB Drives - DB / Backups
>
> Everything is RAID 1+0 except the DB drives. They are RAID 5. However, I am
> wondering if they too should be RAID 1+0. There is always this argument
> about it versus RAID 5.
>
> The databases are not big at all. This is a Departmental DB Server. I have
> approximately 10GB of data and 100MB of Logs. Not large at all.
>
> Please advise.
>
> Thanks.
> -wp
>
>