Hi Everyone
I really need some clarification on this issue before I start using
this new server.
It has 4 Drives in a RAID 10 config and I need to know if I should
partition then for the following:
Setup 1 - everything on C: Partiton
Setup 2 - 2 Partitions
C: OS
D: SQL DATA / SQL LOGFILES
Setup 3 - 3 Partitions
C: OS
D: SQL DATA
E: SQL LOG FILES
I would have thought there would be more head movement in 2 and 3
partitions but I may be wrong.
IU know SQL does use all the memory which is fine it has 4 gig but
what does everyone recommend i do?
MattieUsing logical partitions on the same physical drive array do nothing for
performance but can give the illusion as such to people who may not know
that physical layout. If there is no choice but to have a single drive array
I prefer to have only 1 partition. That way there is never a problem if you
made one too large or too small later down the road.
--
Andrew J. Kelly SQL MVP
"MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
news:1183731888.891067.55660@.g13g2000hsf.googlegroups.com...
> Hi Everyone
> I really need some clarification on this issue before I start using
> this new server.
> It has 4 Drives in a RAID 10 config and I need to know if I should
> partition then for the following:
> Setup 1 - everything on C: Partiton
> Setup 2 - 2 Partitions
> C: OS
> D: SQL DATA / SQL LOGFILES
> Setup 3 - 3 Partitions
> C: OS
> D: SQL DATA
> E: SQL LOG FILES
> I would have thought there would be more head movement in 2 and 3
> partitions but I may be wrong.
> IU know SQL does use all the memory which is fine it has 4 gig but
> what does everyone recommend i do?
> Mattie
>|||MattieG,
I definitely understand Andrew's contention that unpartitioned is less
deceptive. (And partitioning caused me some serious grief once.)
Do you think you will get leverage to eventually separate the disks so that
you have log files on a different physical array? If so, then partitioning
now will allow your file paths to be set up for the future arrival of real
disks. But, if you do not do it now, it is not difficult to move files in
that happy future day.
RLF
"MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
news:1183731888.891067.55660@.g13g2000hsf.googlegroups.com...
> Hi Everyone
> I really need some clarification on this issue before I start using
> this new server.
> It has 4 Drives in a RAID 10 config and I need to know if I should
> partition then for the following:
> Setup 1 - everything on C: Partiton
> Setup 2 - 2 Partitions
> C: OS
> D: SQL DATA / SQL LOGFILES
> Setup 3 - 3 Partitions
> C: OS
> D: SQL DATA
> E: SQL LOG FILES
> I would have thought there would be more head movement in 2 and 3
> partitions but I may be wrong.
> IU know SQL does use all the memory which is fine it has 4 gig but
> what does everyone recommend i do?
> Mattie
>|||Cool, that confirmed my suspicions :)
Im actually looking at getting a couple more disks for the OS or
possibly OS+Database
and maybe use the raid 10 for the log files only.
Unless I use 3 raid 1 arrays for OS / Database / Logfile?
Mattie|||I don't know how busy this will be but if you get two more disks and create
a Raid 1 I would put the OS and possibly the log files (both user and
tempdb) on that array. Then leave the data files on the Raid 10. The OS
doesn't get used much and at least you would separate the data and log
files.
--
Andrew J. Kelly SQL MVP
"MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
news:1183737319.730417.129770@.c77g2000hse.googlegroups.com...
> Cool, that confirmed my suspicions :)
> Im actually looking at getting a couple more disks for the OS or
> possibly OS+Database
> and maybe use the raid 10 for the log files only.
> Unless I use 3 raid 1 arrays for OS / Database / Logfile?
> Mattie
>|||Here is an article I wrote on setting up disk configurations.
http://searchsqlserver.techtarget.com/tip/0,289483,sid87_gci1262122,00.html
--
Denny
MCSA (2003) / MCDBA (SQL 2000)
MCTS (SQL 2005 / Microsoft Windows SharePoint Services 3.0: Configuration /
Microsoft Office SharePoint Server 2007: Configuration)
MCITP (dbadmin, dbdev)
"MattieG" wrote:
> Hi Everyone
> I really need some clarification on this issue before I start using
> this new server.
> It has 4 Drives in a RAID 10 config and I need to know if I should
> partition then for the following:
> Setup 1 - everything on C: Partiton
> Setup 2 - 2 Partitions
> C: OS
> D: SQL DATA / SQL LOGFILES
> Setup 3 - 3 Partitions
> C: OS
> D: SQL DATA
> E: SQL LOG FILES
> I would have thought there would be more head movement in 2 and 3
> partitions but I may be wrong.
> IU know SQL does use all the memory which is fine it has 4 gig but
> what does everyone recommend i do?
> Mattie
>|||?Hi
Thanks for everything :) I will order some more disks.
Mattie|||I will add one comment to Andrew's suggestion. Consider a 4 disk raid 5 set
for your data files if the databases are more select oriented than heavy
insert/update/delete oriented. The 4 disk raid 5 will get you 3 active
spindles for reads which will offer 50% better read throughput than a 4 disk
raid 10 set which has only 2 active spindles serving data. At least I THINK
it works that way. ;)
--
TheSQLGuru
President
Indicium Resources, Inc.
"Andrew J. Kelly" <sqlmvpnooospam@.shadhawk.com> wrote in message
news:O7NP4Q$vHHA.3400@.TK2MSFTNGP03.phx.gbl...
>I don't know how busy this will be but if you get two more disks and create
>a Raid 1 I would put the OS and possibly the log files (both user and
>tempdb) on that array. Then leave the data files on the Raid 10. The OS
>doesn't get used much and at least you would separate the data and log
>files.
> --
> Andrew J. Kelly SQL MVP
> "MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
> news:1183737319.730417.129770@.c77g2000hse.googlegroups.com...
>> Cool, that confirmed my suspicions :)
>> Im actually looking at getting a couple more disks for the OS or
>> possibly OS+Database
>> and maybe use the raid 10 for the log files only.
>> Unless I use 3 raid 1 arrays for OS / Database / Logfile?
>> Mattie
>>
>|||>The 4 disk raid 5 will get you 3 active spindles for reads which will offer
>50% better read throughput than a 4 disk raid 10 set which has only 2
>active spindles serving data.
Not quite. While a 4 disk RAID 5 may outperform a 4 disk Raid 10 for
strictly read operations of large scale it will not be anywhere close to 50%
in my experience. And don't forget that a smart Raid controller can access
both disks in the mirrored pair at the same time for different operations.
--
Andrew J. Kelly SQL MVP
"TheSQLGuru" <kgboles@.earthlink.net> wrote in message
news:eeDOTMkwHHA.600@.TK2MSFTNGP05.phx.gbl...
>I will add one comment to Andrew's suggestion. Consider a 4 disk raid 5
>set for your data files if the databases are more select oriented than
>heavy insert/update/delete oriented. The 4 disk raid 5 will get you 3
>active spindles for reads which will offer 50% better read throughput than
>a 4 disk raid 10 set which has only 2 active spindles serving data. At
>least I THINK it works that way. ;)
> --
> TheSQLGuru
> President
> Indicium Resources, Inc.
> "Andrew J. Kelly" <sqlmvpnooospam@.shadhawk.com> wrote in message
> news:O7NP4Q$vHHA.3400@.TK2MSFTNGP03.phx.gbl...
>>I don't know how busy this will be but if you get two more disks and
>>create a Raid 1 I would put the OS and possibly the log files (both user
>>and tempdb) on that array. Then leave the data files on the Raid 10. The
>>OS doesn't get used much and at least you would separate the data and log
>>files.
>> --
>> Andrew J. Kelly SQL MVP
>> "MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
>> news:1183737319.730417.129770@.c77g2000hse.googlegroups.com...
>> Cool, that confirmed my suspicions :)
>> Im actually looking at getting a couple more disks for the OS or
>> possibly OS+Database
>> and maybe use the raid 10 for the log files only.
>> Unless I use 3 raid 1 arrays for OS / Database / Logfile?
>> Mattie
>>
>>
>|||>>And don't forget that a smart Raid controller can access both disks in
>>the mirrored pair at the same time for different operations.
That was the part I wasn't sure about. I didn't know if the controller
would be able to maintain data integrity if it flipped back and forth
between the drives accessing different areas for different read/write
operations.
TheSQLGuru
President
Indicium Resources, Inc.
"Andrew J. Kelly" <sqlmvpnooospam@.shadhawk.com> wrote in message
news:uapeltnwHHA.4800@.TK2MSFTNGP05.phx.gbl...
> >The 4 disk raid 5 will get you 3 active spindles for reads which will
> >offer 50% better read throughput than a 4 disk raid 10 set which has only
> >2 active spindles serving data.
> Not quite. While a 4 disk RAID 5 may outperform a 4 disk Raid 10 for
> strictly read operations of large scale it will not be anywhere close to
> 50% in my experience. And don't forget that a smart Raid controller can
> access both disks in the mirrored pair at the same time for different
> operations.
> --
> Andrew J. Kelly SQL MVP
> "TheSQLGuru" <kgboles@.earthlink.net> wrote in message
> news:eeDOTMkwHHA.600@.TK2MSFTNGP05.phx.gbl...
>>I will add one comment to Andrew's suggestion. Consider a 4 disk raid 5
>>set for your data files if the databases are more select oriented than
>>heavy insert/update/delete oriented. The 4 disk raid 5 will get you 3
>>active spindles for reads which will offer 50% better read throughput than
>>a 4 disk raid 10 set which has only 2 active spindles serving data. At
>>least I THINK it works that way. ;)
>> --
>> TheSQLGuru
>> President
>> Indicium Resources, Inc.
>> "Andrew J. Kelly" <sqlmvpnooospam@.shadhawk.com> wrote in message
>> news:O7NP4Q$vHHA.3400@.TK2MSFTNGP03.phx.gbl...
>>I don't know how busy this will be but if you get two more disks and
>>create a Raid 1 I would put the OS and possibly the log files (both user
>>and tempdb) on that array. Then leave the data files on the Raid 10. The
>>OS doesn't get used much and at least you would separate the data and log
>>files.
>> --
>> Andrew J. Kelly SQL MVP
>> "MattieG" <mattiegriffin@.hotmail.co.uk> wrote in message
>> news:1183737319.730417.129770@.c77g2000hse.googlegroups.com...
>> Cool, that confirmed my suspicions :)
>> Im actually looking at getting a couple more disks for the OS or
>> possibly OS+Database
>> and maybe use the raid 10 for the log files only.
>> Unless I use 3 raid 1 arrays for OS / Database / Logfile?
>> Mattie
>>
>>
>>
>|||don't think that raid controler is permitted to perform different tasks on
mirrored disks.
smart [very smart & very expensive] raid controler may perform doubled read
from raid1 array, reading at the same time sector_x from disk_0, and
sector_y from disk_1, so having read two sectors at the time tick of just
one.
for this reason raid controler should have very good knowledge of current
head/track location of each disks, and to manage them separately.
information i found just show that this is rarely a case, mostly raid
controler uses just one of raid1 disks for reading, usualy one having better
response time.
the best thing i found useful is using *outer* cylindres for best
performances. on new disks, 25% of outer cylindres has a 25% better
throughoutput then the rest of disk. so, i made on each disk two partitions,
outer, for quick files transfer/read, and inner, for some backup [to not
waste space].
my sql database is only 20gig,, so the smaller outer partition on larger
disks is quite enough.
for tempdb, which needs quick read/write, but not to survive shutdown, maybe
to use i-ram [it is pc card with 4gb ram, connected to mb as ordinarly sata
drive] since it is very fast, but limited to 4gb, possible to form jbod
disk, i-ram and one regular hd, so to allow tempdb, if overgow the 4gb space
to take place on larger second drive.
just my few observations ...
"TheSQLGuru" <kgboles@.earthlink.net> wrote in message
news:uBYpnj8wHHA.3796@.TK2MSFTNGP04.phx.gbl...
> >>And don't forget that a smart Raid controller can access both disks in
> >>the mirrored pair at the same time for different operations.
> That was the part I wasn't sure about. I didn't know if the controller
> would be able to maintain data integrity if it flipped back and forth
> between the drives accessing different areas for different read/write
> operations.
>|||I haven't gone to that level recently but I remember years ago the DPT raid
controllers used to be able to read from one disk and write to the other at
the same time. It would then do the necessary deferred writes to keep them
in sync of coarse. As long as the cache and types of activities were
compatible it seemed to work just fine. So I assume with more modern
technology this is still true but I can't confirm that for sure.
--
Andrew J. Kelly SQL MVP
"sali" <gabor.salai@.tel.net.ba> wrote in message
news:%236csKJ$wHHA.1208@.TK2MSFTNGP03.phx.gbl...
> don't think that raid controler is permitted to perform different tasks on
> mirrored disks.
> smart [very smart & very expensive] raid controler may perform doubled
> read
> from raid1 array, reading at the same time sector_x from disk_0, and
> sector_y from disk_1, so having read two sectors at the time tick of just
> one.
> for this reason raid controler should have very good knowledge of current
> head/track location of each disks, and to manage them separately.
> information i found just show that this is rarely a case, mostly raid
> controler uses just one of raid1 disks for reading, usualy one having
> better
> response time.
> the best thing i found useful is using *outer* cylindres for best
> performances. on new disks, 25% of outer cylindres has a 25% better
> throughoutput then the rest of disk. so, i made on each disk two
> partitions,
> outer, for quick files transfer/read, and inner, for some backup [to not
> waste space].
> my sql database is only 20gig,, so the smaller outer partition on larger
> disks is quite enough.
> for tempdb, which needs quick read/write, but not to survive shutdown,
> maybe
> to use i-ram [it is pc card with 4gb ram, connected to mb as ordinarly
> sata
> drive] since it is very fast, but limited to 4gb, possible to form jbod
> disk, i-ram and one regular hd, so to allow tempdb, if overgow the 4gb
> space
> to take place on larger second drive.
> just my few observations ...
>
> "TheSQLGuru" <kgboles@.earthlink.net> wrote in message
> news:uBYpnj8wHHA.3796@.TK2MSFTNGP04.phx.gbl...
>> >>And don't forget that a smart Raid controller can access both disks in
>> >>the mirrored pair at the same time for different operations.
>> That was the part I wasn't sure about. I didn't know if the controller
>> would be able to maintain data integrity if it flipped back and forth
>> between the drives accessing different areas for different read/write
>> operations.
>
>
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment