| Author |
Message |
Guest
|
 Posted:
Fri Sep 23, 2005 1:35 pm Post subject:
RAID1 performance vs single disk configuration |
|
|
Hello,
how much better is the performance of RAID1 (U320 SCSI)
vs a single disk configuration in a db-environment (only
as a very GENERAL statement) ?
Any tip is appreciated. Thank`s a lot!
John |
|
| Back to top |
|
 |
Jeremy Boden
Guest
|
| Posted:
Fri Sep 23, 2005 4:18 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
In message <1127464503.258624.137140@g43g2000cwa.googlegroups.com>,
hardware55@bluemail.ch writes
| Quote: | Hello,
how much better is the performance of RAID1 (U320 SCSI)
vs a single disk configuration in a db-environment (only
as a very GENERAL statement) ?
Any tip is appreciated. Thank`s a lot!
John
By "db-environment" I assume you mean random (non-sequential) access? |
Note that RAID 1 requires a minimum of two drives, so your comparison
should be with two single drives. So, especially if you are
multi-tasking etc etc then the non-RAID solution ought to be faster -
especially if you have your databases "nicely" distributed.
At least until a disk fails.
--
Jeremy Boden |
|
| Back to top |
|
|
Eric Gisin
Guest
|
| Posted:
Fri Sep 23, 2005 7:53 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
Xbitlabs.com has some ScSI RAiD benchmarks.
<hardware55@bluemail.ch> wrote in message
news:1127464503.258624.137140@g43g2000cwa.googlegroups.com...
| Quote: |
how much better is the performance of RAID1 (U320 SCSI)
vs a single disk configuration in a db-environment (only
as a very GENERAL statement) ?
|
|
|
| Back to top |
|
|
Curious George
Guest
|
| Posted:
Mon Oct 03, 2005 7:26 am Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
On Fri, 23 Sep 2005 12:18:00 +0100, Jeremy Boden
<jeremy@jboden.demon.co.uk> wrote:
| Quote: | In message <1127464503.258624.137140@g43g2000cwa.googlegroups.com>,
hardware55@bluemail.ch writes
Hello,
how much better is the performance of RAID1 (U320 SCSI)
vs a single disk configuration in a db-environment (only
as a very GENERAL statement) ?
Any tip is appreciated. Thank`s a lot!
John
By "db-environment" I assume you mean random (non-sequential) access?
Note that RAID 1 requires a minimum of two drives,
|
Note that RAID 1 requires exactly 2 drives
| Quote: | so your comparison
should be with two single drives.
|
1 volume (single disk) to 1 volume (raid 1) is a very reasonable
comparison. Not all raid 1 are created equal so he can expect between
1x to theoretically 2x performance of a single disk in certain
situations & HW.
| Quote: | So, especially if you are
multi-tasking etc etc then the non-RAID solution ought to be faster -
especially if you have your databases "nicely" distributed.
|
Distributing the workload is a fine idea when & to whatever extent
possible. Often the distribution can't be "nice" or at least it can't
make up for insufficiently performing storage for even specialized
jobs. that's when other raid levels start getting looked at closely.
| Quote: | At least until a disk fails.
|
IMHO your last two points make a better argument for >1 array rather
than merely >1 disk in certain machines. |
|
| Back to top |
|
|
Ray
Guest
|
| Posted:
Mon Oct 03, 2005 10:23 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
| Quote: | Note that RAID 1 requires a minimum of two drives,
Note that RAID 1 requires exactly 2 drives
|
The first statement is correct. The world is not Microsoft. There are
vendors who offer mirroring across more that 2 drives.
For example, even an old HP HSC50 controller can manage up to 24 disks in
one mirror set. |
|
| Back to top |
|
|
Folkert Rienstra
Guest
|
| Posted:
Mon Oct 03, 2005 11:06 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
"Ray" <no@spam.me> wrote in message news:gOd0f.4013$Yv6.3761@fe06.lga
| Quote: | Note that RAID 1 requires a minimum of two drives,
Note that RAID 1 requires exactly 2 drives
The first statement is correct. The world is not Microsoft.
|
Has nothing to do with Microsoft.
| Quote: | There are vendors who offer mirroring across more that 2 drives.
|
What makes you think Microsoft can't do that too?
| Quote: |
For example, even an old HP HSC50 controller can manage up to 24 disks in
one mirror set.
|
But it isn't RAID1. |
|
| Back to top |
|
|
Folkert Rienstra
Guest
|
| Posted:
Mon Oct 03, 2005 11:15 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
"Eric Gisin" <ericgisin@hotmail.com> wrote in message news:dh15fg018j4@enews4.newsguy.com
| Quote: | In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
|
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
| Quote: |
Xbitlabs.com has some ScSI RAiD benchmarks.
hardware55@bluemail.ch> wrote in message news:1127464503.258624.137140@g43g2000cwa.googlegroups.com...
how much better is the performance of RAID1 (U320 SCSI)
vs a single disk configuration in a db-environment (only
as a very GENERAL statement) ? |
|
|
| Back to top |
|
|
Eric Gisin
Guest
|
| Posted:
Tue Oct 04, 2005 4:31 am Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
"Folkert Rienstra" <see_reply-to@myweb.nl> wrote in message
news:4341b5de$1$76595$892e7fe2@authen.white.readfreenews.net...
| Quote: | "Eric Gisin" <ericgisin@hotmail.com> wrote in message news:dh15fg018j4@enews4.newsguy.com
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
The OS knows the last LBA on each drive. |
Select the one closest to the requested LBA. |
|
| Back to top |
|
|
Jeremy Boden
Guest
|
| Posted:
Tue Oct 04, 2005 4:56 am Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
In message <4341b5de$1$76595$892e7fe2@authen.white.readfreenews.net>,
Folkert Rienstra <see_reply-to@myweb.nl> writes
| Quote: | "Eric Gisin" <ericgisin@hotmail.com> wrote in message
news:dh15fg018j4@enews4.newsguy.com
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
..... |
Then what is the point in queuing requests to targets?
--
Jeremy Boden |
|
| Back to top |
|
|
Folkert Rienstra
Guest
|
| Posted:
Wed Oct 05, 2005 12:18 am Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
"Eric Gisin" <ericgisin@hotmail.com> wrote in message news:dhsfi401fkb@enews2.newsguy.com...
| Quote: | "Folkert Rienstra" <see_reply-to@myweb.nl> wrote in message news:4341b5de$1$76595$892e7fe2@authen.white.readfreenews.net...
"Eric Gisin" <ericgisin@hotmail.com> wrote in message news:dh15fg018j4@enews4.newsguy.com
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
The OS knows the last LBA on each drive.
|
But not the rotational position ...
| Quote: | Select the one closest to the requested LBA.
|
.... which may well cancel the effort if that drive just passed the
sector and the drive furthest away might have (still) got it in time.
|
|
| Back to top |
|
|
Curious George
Guest
|
| Posted:
Wed Oct 05, 2005 7:26 am Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
On Mon, 3 Oct 2005 13:23:36 -0400, "Ray" <no@spam.me> wrote:
| Quote: | Note that RAID 1 requires a minimum of two drives,
Note that RAID 1 requires exactly 2 drives
The first statement is correct. The world is not Microsoft. There are
vendors who offer mirroring across more that 2 drives.
For example, even an old HP HSC50 controller can manage up to 24 disks in
one mirror set.
|
Those are compund levels, not simple RAID1 |
|
| Back to top |
|
|
Jeremy Boden
Guest
|
| Posted:
Wed Oct 05, 2005 2:11 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
In message <4342d5a9$0$4754$892e7fe2@authen.white.readfreenews.net>,
Folkert Rienstra <see_reply-to@myweb.nl> writes
| Quote: | "Eric Gisin" <ericgisin@hotmail.com> wrote in message
news:dhsfi401fkb@enews2.newsguy.com...
"Folkert Rienstra" <see_reply-to@myweb.nl> wrote in message
news:4341b5de$1$76595$892e7fe2@authen.white.readfreenews.net...
"Eric Gisin" <ericgisin@hotmail.com> wrote in message
news:dh15fg018j4@enews4.newsguy.com
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
The OS knows the last LBA on each drive.
But not the rotational position ...
Select the one closest to the requested LBA.
... which may well cancel the effort if that drive just passed the
sector and the drive furthest away might have (still) got it in time.
Always assuming both heads were aligned on the same track... |
--
Jeremy Boden |
|
| Back to top |
|
|
Curious George
Guest
|
| Posted:
Wed Oct 05, 2005 11:18 pm Post subject:
Re: RAID1 performance vs single disk configuration |
|
|
On Wed, 5 Oct 2005 10:11:59 +0100, Jeremy Boden
<jeremy@jboden.demon.co.uk> wrote:
| Quote: | In message <4342d5a9$0$4754$892e7fe2@authen.white.readfreenews.net>,
Folkert Rienstra <see_reply-to@myweb.nl> writes
"Eric Gisin" <ericgisin@hotmail.com> wrote in message
news:dhsfi401fkb@enews2.newsguy.com...
"Folkert Rienstra" <see_reply-to@myweb.nl> wrote in message
news:4341b5de$1$76595$892e7fe2@authen.white.readfreenews.net...
"Eric Gisin" <ericgisin@hotmail.com> wrote in message
news:dh15fg018j4@enews4.newsguy.com
In read-only server testing, RAID 1 is better than RAID 0 or JBOD.
That's because read requests can be assigned to either drive.
I would assume that the performance is same in *theory*.
In practice both will suffer from different sized transfers.
Better latency may make the difference for RAID1 but I fail to see
how the OS or driver can predict which drive will have the data first.
The OS knows the last LBA on each drive.
But not the rotational position ...
Select the one closest to the requested LBA.
... which may well cancel the effort if that drive just passed the
sector and the drive furthest away might have (still) got it in time.
Always assuming both heads were aligned on the same track...
|
You shouldn't. RPL has been obsolete for years. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in