md on partition

Hi folks.

I remembered that there was a discussion on the list about some abnormal
performance penalty if u use a partition instead of a whole disk for MD.
But I could not find this discussion via archival or google.

Could somebody be kindly enough to point me to the right spot? Thanks a
lot.

Ming


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

I found this old message:
"On Tuesday June 1, maheshext3@yahoo.com wrote:
>
> --- Neil Brown wrote:
> >
> > What does
> > ls -l /dev/hd[eg]5
> > show? How about
> > cat /proc/partitions
> > dd if=/dev/hde5 of=/dev/null bs=1024k
>
> Thanks a lot, Neil, turns out those nodes were not present. After I
> mknod'd them, all's fine in RAIDland...
> Quick question: does using extended partitions for RAID affect
> performance?

The fact that the partitions are "extended" is invisible to most of the
kernel, and it could not have any affect of performance.

NeilBrown"

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
Sent: Friday, October 22, 2004 1:28 PM
To: linux-raid@vger.kernel.org
Subject: md on partition

Hi folks.

I remembered that there was a discussion on the list about some abnormal
performance penalty if u use a partition instead of a whole disk for MD.
But I could not find this discussion via archival or google.

Could somebody be kindly enough to point me to the right spot? Thanks a
lot.

Ming


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

thx but i met this problem.

i use 4 x 400GB SATA disks, i make a raid 0 or 5, and build a vg on top
of md, then use bonnie++ to test read performance.

LVM (MD = sda+sdb+..) = 95MBsec, md is built on top of whole disk.

LVM (MD = sda1+sdb1+.) = 48MBsec, md is built on top of each partition.

so where is the problem? does md chunk size matter? or lvm extent size
matters? thx.


ming

On Sat, 2004-10-23 at 02:07, Guy wrote:
> I found this old message:
> "On Tuesday June 1, maheshext3@yahoo.com wrote:
> >
> > --- Neil Brown wrote:
> > >
> > > What does
> > > ls -l /dev/hd[eg]5
> > > show? How about
> > > cat /proc/partitions
> > > dd if=/dev/hde5 of=/dev/null bs=1024k
> >
> > Thanks a lot, Neil, turns out those nodes were not present. After I
> > mknod'd them, all's fine in RAIDland...
> > Quick question: does using extended partitions for RAID affect
> > performance?
>
> The fact that the partitions are "extended" is invisible to most of the
> kernel, and it could not have any affect of performance.
>
> NeilBrown"
>
> Guy
>
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> Sent: Friday, October 22, 2004 1:28 PM
> To: linux-raid@vger.kernel.org
> Subject: md on partition
>
> Hi folks.
>
> I remembered that there was a discussion on the list about some abnormal
> performance penalty if u use a partition instead of a whole disk for MD.
> But I could not find this discussion via archival or google.
>
> Could somebody be kindly enough to point me to the right spot? Thanks a
> lot.
>
> Ming
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
| http://crab.ele.uri.edu/gallery/albums.php
--------------------------------------------------


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

I guess Neil was wrong, or I misunderstood something.
Were both test with the same RAID and LVM config? Or was 1 RAID0 and the
other RAID5?

If RAID5, are you sure the resync was done before you did your tests?
"cat /proc/mdstat" to determine this.

Yes, size does matter! Chunk size that is. Not sure about extent size.

As long as the only difference was the whole disk vs partitions, then
partitions make a big difference! This sucks!

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
Sent: Saturday, October 23, 2004 11:52 AM
To: Guy
Cc: linux-raid@vger.kernel.org
Subject: RE: md on partition

thx but i met this problem.

i use 4 x 400GB SATA disks, i make a raid 0 or 5, and build a vg on top
of md, then use bonnie++ to test read performance.

LVM (MD = sda+sdb+..) = 95MBsec, md is built on top of whole disk.

LVM (MD = sda1+sdb1+.) = 48MBsec, md is built on top of each partition.

so where is the problem? does md chunk size matter? or lvm extent size
matters? thx.


ming

On Sat, 2004-10-23 at 02:07, Guy wrote:
> I found this old message:
> "On Tuesday June 1, maheshext3@yahoo.com wrote:
> >
> > --- Neil Brown wrote:
> > >
> > > What does
> > > ls -l /dev/hd[eg]5
> > > show? How about
> > > cat /proc/partitions
> > > dd if=/dev/hde5 of=/dev/null bs=1024k
> >
> > Thanks a lot, Neil, turns out those nodes were not present. After I
> > mknod'd them, all's fine in RAIDland...
> > Quick question: does using extended partitions for RAID affect
> > performance?
>
> The fact that the partitions are "extended" is invisible to most of the
> kernel, and it could not have any affect of performance.
>
> NeilBrown"
>
> Guy
>
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> Sent: Friday, October 22, 2004 1:28 PM
> To: linux-raid@vger.kernel.org
> Subject: md on partition
>
> Hi folks.
>
> I remembered that there was a discussion on the list about some abnormal
> performance penalty if u use a partition instead of a whole disk for MD.
> But I could not find this discussion via archival or google.
>
> Could somebody be kindly enough to point me to the right spot? Thanks a
> lot.
>
> Ming
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
| http://crab.ele.uri.edu/gallery/albums.php
--------------------------------------------------


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

On Sat, 2004-10-23 at 12:12, Guy wrote:
> I guess Neil was wrong, or I misunderstood something.
> Were both test with the same RAID and LVM config? Or was 1 RAID0 and the
> other RAID5?
yes, if i tested RAID0, then will use RAID0 for both partition or whole
disk scenarios.

>
> If RAID5, are you sure the resync was done before you did your tests?
> "cat /proc/mdstat" to determine this.
>
need to recheck. :)

> Yes, size does matter! Chunk size that is. Not sure about extent size.
>
will raid by default choose different chunk size if device size is
different, for example, sda is 400GB while sda1 is 50GB?

> As long as the only difference was the whole disk vs partitions, then
> partitions make a big difference! This sucks!
>
> Guy
>
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> Sent: Saturday, October 23, 2004 11:52 AM
> To: Guy
> Cc: linux-raid@vger.kernel.org
> Subject: RE: md on partition
>
> thx but i met this problem.
>
> i use 4 x 400GB SATA disks, i make a raid 0 or 5, and build a vg on top
> of md, then use bonnie++ to test read performance.
>
> LVM (MD = sda+sdb+..) = 95MBsec, md is built on top of whole disk.
>
> LVM (MD = sda1+sdb1+.) = 48MBsec, md is built on top of each partition.
>
> so where is the problem? does md chunk size matter? or lvm extent size
> matters? thx.
>
>
> ming
>
> On Sat, 2004-10-23 at 02:07, Guy wrote:
> > I found this old message:
> > "On Tuesday June 1, maheshext3@yahoo.com wrote:
> > >
> > > --- Neil Brown wrote:
> > > >
> > > > What does
> > > > ls -l /dev/hd[eg]5
> > > > show? How about
> > > > cat /proc/partitions
> > > > dd if=/dev/hde5 of=/dev/null bs=1024k
> > >
> > > Thanks a lot, Neil, turns out those nodes were not present. After I
> > > mknod'd them, all's fine in RAIDland...
> > > Quick question: does using extended partitions for RAID affect
> > > performance?
> >
> > The fact that the partitions are "extended" is invisible to most of the
> > kernel, and it could not have any affect of performance.
> >
> > NeilBrown"
> >
> > Guy
> >
> > -----Original Message-----
> > From: linux-raid-owner@vger.kernel.org
> > [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> > Sent: Friday, October 22, 2004 1:28 PM
> > To: linux-raid@vger.kernel.org
> > Subject: md on partition
> >
> > Hi folks.
> >
> > I remembered that there was a discussion on the list about some abnormal
> > performance penalty if u use a partition instead of a whole disk for MD.
> > But I could not find this discussion via archival or google.
> >
> > Could somebody be kindly enough to point me to the right spot? Thanks a
> > lot.
> >
> > Ming
> >
> >
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
| http://crab.ele.uri.edu/gallery/albums.php
--------------------------------------------------


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

You said:
"will raid by default choose different chunk size if device size is
different, for example, sda is 400GB while sda1 is 50GB?"

No, but the random seek test will give different results. 50GB will perform
better, since less head travel.

Why the difference? I would want everything the same for both tests.

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
Sent: Saturday, October 23, 2004 6:39 PM
To: Guy
Cc: linux-raid@vger.kernel.org
Subject: RE: md on partition

On Sat, 2004-10-23 at 12:12, Guy wrote:
> I guess Neil was wrong, or I misunderstood something.
> Were both test with the same RAID and LVM config? Or was 1 RAID0 and the
> other RAID5?
yes, if i tested RAID0, then will use RAID0 for both partition or whole
disk scenarios.

>
> If RAID5, are you sure the resync was done before you did your tests?
> "cat /proc/mdstat" to determine this.
>
need to recheck. :)

> Yes, size does matter! Chunk size that is. Not sure about extent size.
>
will raid by default choose different chunk size if device size is
different, for example, sda is 400GB while sda1 is 50GB?

> As long as the only difference was the whole disk vs partitions, then
> partitions make a big difference! This sucks!
>
> Guy
>
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> Sent: Saturday, October 23, 2004 11:52 AM
> To: Guy
> Cc: linux-raid@vger.kernel.org
> Subject: RE: md on partition
>
> thx but i met this problem.
>
> i use 4 x 400GB SATA disks, i make a raid 0 or 5, and build a vg on top
> of md, then use bonnie++ to test read performance.
>
> LVM (MD = sda+sdb+..) = 95MBsec, md is built on top of whole disk.
>
> LVM (MD = sda1+sdb1+.) = 48MBsec, md is built on top of each partition.
>
> so where is the problem? does md chunk size matter? or lvm extent size
> matters? thx.
>
>
> ming
>
> On Sat, 2004-10-23 at 02:07, Guy wrote:
> > I found this old message:
> > "On Tuesday June 1, maheshext3@yahoo.com wrote:
> > >
> > > --- Neil Brown wrote:
> > > >
> > > > What does
> > > > ls -l /dev/hd[eg]5
> > > > show? How about
> > > > cat /proc/partitions
> > > > dd if=/dev/hde5 of=/dev/null bs=1024k
> > >
> > > Thanks a lot, Neil, turns out those nodes were not present. After I
> > > mknod'd them, all's fine in RAIDland...
> > > Quick question: does using extended partitions for RAID affect
> > > performance?
> >
> > The fact that the partitions are "extended" is invisible to most of the
> > kernel, and it could not have any affect of performance.
> >
> > NeilBrown"
> >
> > Guy
> >
> > -----Original Message-----
> > From: linux-raid-owner@vger.kernel.org
> > [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> > Sent: Friday, October 22, 2004 1:28 PM
> > To: linux-raid@vger.kernel.org
> > Subject: md on partition
> >
> > Hi folks.
> >
> > I remembered that there was a discussion on the list about some abnormal
> > performance penalty if u use a partition instead of a whole disk for MD.
> > But I could not find this discussion via archival or google.
> >
> > Could somebody be kindly enough to point me to the right spot? Thanks a
> > lot.
> >
> > Ming
> >
> >
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
| http://crab.ele.uri.edu/gallery/albums.php
--------------------------------------------------


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: md on partition

On Sat, 2004-10-23 at 18:52, Guy wrote:
> You said:
> "will raid by default choose different chunk size if device size is
> different, for example, sda is 400GB while sda1 is 50GB?"
>
> No, but the random seek test will give different results. 50GB will perform
> better, since less head travel.
>
> Why the difference? I would want everything the same for both tests.
>
so assume raid use same chunk size for both 50GB partition and 400GB
disk, i can not understand why here a near half performance down grade.

> Guy
>
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> Sent: Saturday, October 23, 2004 6:39 PM
> To: Guy
> Cc: linux-raid@vger.kernel.org
> Subject: RE: md on partition
>
> On Sat, 2004-10-23 at 12:12, Guy wrote:
> > I guess Neil was wrong, or I misunderstood something.
> > Were both test with the same RAID and LVM config? Or was 1 RAID0 and the
> > other RAID5?
> yes, if i tested RAID0, then will use RAID0 for both partition or whole
> disk scenarios.
>
> >
> > If RAID5, are you sure the resync was done before you did your tests?
> > "cat /proc/mdstat" to determine this.
> >
> need to recheck. :)
>
> > Yes, size does matter! Chunk size that is. Not sure about extent size.
> >
> will raid by default choose different chunk size if device size is
> different, for example, sda is 400GB while sda1 is 50GB?
>
> > As long as the only difference was the whole disk vs partitions, then
> > partitions make a big difference! This sucks!
> >
> > Guy
> >
> > -----Original Message-----
> > From: linux-raid-owner@vger.kernel.org
> > [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> > Sent: Saturday, October 23, 2004 11:52 AM
> > To: Guy
> > Cc: linux-raid@vger.kernel.org
> > Subject: RE: md on partition
> >
> > thx but i met this problem.
> >
> > i use 4 x 400GB SATA disks, i make a raid 0 or 5, and build a vg on top
> > of md, then use bonnie++ to test read performance.
> >
> > LVM (MD = sda+sdb+..) = 95MBsec, md is built on top of whole disk.
> >
> > LVM (MD = sda1+sdb1+.) = 48MBsec, md is built on top of each partition.
> >
> > so where is the problem? does md chunk size matter? or lvm extent size
> > matters? thx.
> >
> >
> > ming
> >
> > On Sat, 2004-10-23 at 02:07, Guy wrote:
> > > I found this old message:
> > > "On Tuesday June 1, maheshext3@yahoo.com wrote:
> > > >
> > > > --- Neil Brown wrote:
> > > > >
> > > > > What does
> > > > > ls -l /dev/hd[eg]5
> > > > > show? How about
> > > > > cat /proc/partitions
> > > > > dd if=/dev/hde5 of=/dev/null bs=1024k
> > > >
> > > > Thanks a lot, Neil, turns out those nodes were not present. After I
> > > > mknod'd them, all's fine in RAIDland...
> > > > Quick question: does using extended partitions for RAID affect
> > > > performance?
> > >
> > > The fact that the partitions are "extended" is invisible to most of the
> > > kernel, and it could not have any affect of performance.
> > >
> > > NeilBrown"
> > >
> > > Guy
> > >
> > > -----Original Message-----
> > > From: linux-raid-owner@vger.kernel.org
> > > [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ming Zhang
> > > Sent: Friday, October 22, 2004 1:28 PM
> > > To: linux-raid@vger.kernel.org
> > > Subject: md on partition
> > >
> > > Hi folks.
> > >
> > > I remembered that there was a discussion on the list about some abnormal
> > > performance penalty if u use a partition instead of a whole disk for MD.
> > > But I could not find this discussion via archival or google.
> > >
> > > Could somebody be kindly enough to point me to the right spot? Thanks a
> > > lot.
> > >
> > > Ming
> > >
> > >
> > > -
> > > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at http://vger.kernel.org/majordomo-info.html
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
| http://crab.ele.uri.edu/gallery/albums.php
--------------------------------------------------


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html