GNU bug report logs -
#22180
Partition alignment using parted
Previous Next
To reply to this bug, email your comments to 22180 AT debbugs.gnu.org.
Toggle the display of automated, internal messages from the tracker.
Report forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Tue, 15 Dec 2015 16:56:02 GMT)
Full text and
rfc822 format available.
Acknowledgement sent
to
Ankur Tank <Ankur.Tank <at> LntTechservices.com>:
New bug report received and forwarded. Copy sent to
bug-parted <at> gnu.org.
(Tue, 15 Dec 2015 16:56:02 GMT)
Full text and
rfc822 format available.
Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):
Hi Experts,
I used following in my script,
I am not sure if below behavior is bug or expected behavior Just to confirm,
parted --script -a optimal /dev/mmcblk0 \
mklabel gpt \
mkpart primary 0% 219s \
mkpart primary 220s 475s \
mkpart primary 476s 731s \
mkpart primary 732s 987s \
mkpart primary 988s 3035s \
mkpart primary 3036s 5083s \
mkpart primary 5084s 6107s \
mkpart primary 6108s 6363s \
mkpart primary 6364s 16603s \
mkpart primary 16604s 26843s \
mkpart primary 26844s 37083s \
mkpart primary 37084s 37595s \
mkpart primary 37596s 38107s \
mkpart primary 38108s 38619s \
mkpart primary 38620s 73403s \
mkpart primary 73404s 483003s \
mkpart primary 483004s 892603s \
mkpart primary 892604s 1097403s \
mkpart primary 1097404s 1220283s \
mkpart primary 1220284s 1629883s \
mkpart primary 1629884s 1701563s \
mkpart primary 1701564s 100%
But I see following message from parted,
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
When I list partitions I see following values
(---------Last column I added to inform that I expected which size size--------)
# parted --list
Model: MMC MMC04G (sd/mmc)
Disk /dev/mmcblk0: 3842MB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Disk Flags:
Number Start End Size File system Name Flags Expected Size(in KB)
1 17.4kB 113kB 95.2kB primary 110
2 113kB 244kB 131kB primary 128
3 244kB 375kB 131kB primary 128
4 375kB 506kB 131kB primary 128
5 506kB 1554kB 1049kB primary 1024
6 1554kB 2603kB 1049kB primary 1024
7 2603kB 3127kB 524kB primary 512
8 3127kB 3258kB 131kB primary 128
9 3258kB 8501kB 5243kB primary 5120
10 8501kB 13.7MB 5243kB primary 5120
11 13.7MB 19.0MB 5243kB primary 5120
12 19.0MB 19.2MB 262kB primary 256
13 19.2MB 19.5MB 262kB primary 256
14 19.5MB 19.8MB 262kB primary 256
15 19.8MB 37.6MB 17.8MB primary 17392
16 37.6MB 247MB 210MB primary 204800
17 247MB 457MB 210MB primary 204800
18 457MB 562MB 105MB primary 102400
19 562MB 625MB 62.9MB primary 61440
20 625MB 835MB 210MB primary 204800
21 835MB 871MB 36.7MB primary 35840
22 871MB 3842MB 2971MB primary 2969600
Could someone explain why parted is not giving partition size which I expected ? Because I had aligned partitions to sector boundaries(sector size 512B).
Any suggestions/pointers/corrections ?
Thank you,
Regards,
Ankur
-----Original Message-----
From: Ankur Tank
Sent: Tuesday, December 15, 2015 3:56 PM
To: 'bug-parted <at> gnu.org'
Cc: artfri2 <at> gmail.com
Subject: Partition alignment using parted
Hi Parted Experts,
I am working on am335x based board.
We are using eMMC and as a storage medium and parted for partitioning eMMC.
I am not able to use "optimal" alignment option with parted when partitioning eMMC.
I get following error messages when I use "optimal" alignment
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
Warning: The resulting partition is not properly aligned for best performance.
1. Do we have to worry about this message ?
I have put question on Unix & Linux Stackexchange also below is the link for the same http://unix.stackexchange.com/questions/248939/how-to-achieve-optimal-alignment-for-emmc-partition
Now parted reports logical and physical sector size as same.
"Sector size (logical/physical): 512B/512B"
2. Now my question is, is physical size reported by parted is physical sector size or emulated physical sector size?
3. How can we achieve optimal alignment ? Any suggestions?
Thank you,
Regards,
Ankur
L&T Technology Services Ltd
www.LntTechservices.com<http://www.lnttechservices.com/>
This Email may contain confidential or privileged information for the intended recipient (s). If you are not the intended recipient, please do not use or disseminate the information, notify the sender and delete it from your system.
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Wed, 23 Dec 2015 01:11:02 GMT)
Full text and
rfc822 format available.
Message #8 received at 22180 <at> debbugs.gnu.org (full text, mbox):
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Control: forcemerge -1 22179
Control: close -1
Partitions can not possibly be created on a fraction of a sector.
Optimal alignment is much larger than a single sector: typically 1
MiB. Create your partitions on 1 MiB boundaries. When you specify
the size in bytes or sectors, you are asking for exact placement.
Using conventional units of MB or GB allows parted to move the exact
start and end to the nearest preferable location rather than giving
that warning.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAEBCgAGBQJWefRxAAoJEBB5UWFcu6UWOEYIAIK+AQhtBKdR8E26kRx+mTrW
ugsZzi2dQMfJ4ECYGBbMMLoXdFr2KSNaNZe6MjESvOxYKCZW6NZMKMeMdvpQNaXq
rHClHg+ra8JNo7QFqGfAn2BWYvbXfnMtmmrUdGRQZQBI8h7KZiIG6gts/GD9uKEO
puqmJEpRku4TQFR1YOEKAJDGlb9DhoKKlTAXInb2Ctzk1RzsP+D1J6DAjGQwPu5x
/L8A7BEuwigq9eFA0KdxCNTzcOv/et+ddwCf600WrbToqeWSSAf+LuMmiYbK2TKH
KEp8lnZfvOZAIv7ogCTNG4X1Dbc8tyEMpOW9MsKp7cOWUjbtDaonaMsNpu+n2AM=
=DBsF
-----END PGP SIGNATURE-----
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Wed, 23 Dec 2015 17:20:01 GMT)
Full text and
rfc822 format available.
Message #11 received at 22180 <at> debbugs.gnu.org (full text, mbox):
On 12/23/2015 8:35 AM, Ankur Tank wrote:
> Hi Phillip
>
> Thank you very much for reply, We are using am335x based (Beaglebone
> Black) custom board with eMMC(4GB). I have couple more questions
> regarding partition alignment. 1. Does partition alignment guarantee
> performance improvement ?
No; it depends entirely on the underlying storage device. Your average
hard disk does not care at all about alignment. Hard disks that
internally use 4k physical sectors but report 512 byte logical sectors
will see a significant performance difference between a partition that
is aligned to 4k or not. A raid array performs much better when aligned
to a multiple of the stripe size. The performance characteristics of
flash vary greatly.
> 2. Is it required for all the partitions to be aligned ? Because we
> are creating four raw partitions at the beginning of the eMMC which
> can't be on 1MB boundary.(because ROM code looks at on 128kb blocks
> only).
No, it is not required. And especially for a mostly read only partition,
it really doesn't matter as it is only writes that suffer from improper
alignment.
> 3. Is there any better way to get partition aligned. (I know I
> sound silly but I wanted to know if there is any better method ).
I'm not sure what you mean.
> 4.
> parted is not allowing to use ~340 MB of size at the end of the eMMC,
> why so ? Am I missing something ? 5. I don't see verbose option for
> parted in man page, Is there any way to enable it?
I can't read your mind, so you will have to describe what you are
actually seeing.
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Wed, 23 Dec 2015 17:40:02 GMT)
Full text and
rfc822 format available.
Message #14 received at 22180 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
Hi Phillip
Thank you very much for reply,
We are using am335x based (Beaglebone Black) custom board with eMMC(4GB).
I have couple more questions regarding partition alignment.
1. Does partition alignment guarantee performance improvement ?
2. Is it required for all the partitions to be aligned ? Because we are creating four raw partitions at the beginning of the eMMC which can't be on 1MB boundary.(because ROM code looks at on 128kb blocks only).
3. Is there any better way to get partition aligned. (I know I sound silly but I wanted to know if there is any better method ).
4. parted is not allowing to use ~340 MB of size at the end of the eMMC, why so ? Am I missing something ?
5. I don't see verbose option for parted in man page, Is there any way to enable it?
I know its atrocious to ask you to look at my partition script and partition calculator excel sheet.
However if you can review them, it would be a great help, I have attached them with this mail.
Any pointers/suggestions ?
Thank you,
Regards,
Ankur
-----Original Message-----
From: Phillip Susi [mailto:psusi <at> ubuntu.com]
Sent: Wednesday, December 23, 2015 6:40 AM
To: Ankur Tank; 22180 <at> debbugs.gnu.org
Cc: artfri2 <at> gmail.com
Subject: Re: bug#22180: Partition alignment using parted
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Control: forcemerge -1 22179
Control: close -1
Partitions can not possibly be created on a fraction of a sector.
Optimal alignment is much larger than a single sector: typically 1 MiB. Create your partitions on 1 MiB boundaries. When you specify the size in bytes or sectors, you are asking for exact placement.
Using conventional units of MB or GB allows parted to move the exact start and end to the nearest preferable location rather than giving that warning.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAEBCgAGBQJWefRxAAoJEBB5UWFcu6UWOEYIAIK+AQhtBKdR8E26kRx+mTrW
ugsZzi2dQMfJ4ECYGBbMMLoXdFr2KSNaNZe6MjESvOxYKCZW6NZMKMeMdvpQNaXq
rHClHg+ra8JNo7QFqGfAn2BWYvbXfnMtmmrUdGRQZQBI8h7KZiIG6gts/GD9uKEO
puqmJEpRku4TQFR1YOEKAJDGlb9DhoKKlTAXInb2Ctzk1RzsP+D1J6DAjGQwPu5x
/L8A7BEuwigq9eFA0KdxCNTzcOv/et+ddwCf600WrbToqeWSSAf+LuMmiYbK2TKH
KEp8lnZfvOZAIv7ogCTNG4X1Dbc8tyEMpOW9MsKp7cOWUjbtDaonaMsNpu+n2AM=
=DBsF
-----END PGP SIGNATURE-----
L&T Technology Services Ltd
www.LntTechservices.com<http://www.lnttechservices.com/>
This Email may contain confidential or privileged information for the intended recipient (s). If you are not the intended recipient, please do not use or disseminate the information, notify the sender and delete it from your system.
[partition_emmc.tar.gz (application/x-gzip, attachment)]
[PartitionCalculator_2.xlsx (application/vnd.openxmlformats-officedocument.spreadsheetml.sheet, attachment)]
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Thu, 24 Dec 2015 10:52:01 GMT)
Full text and
rfc822 format available.
Message #17 received at 22180 <at> debbugs.gnu.org (full text, mbox):
Thank you for replying Phillip,
1. How do we measure performance improvement before and after partition alignment ?
> 3. Is there any better way to get partition aligned. (I know I sound
> silly but I wanted to know if there is any better method ).
2. By above question, I meant, are there any tools/utilities available which would suggest partition alignment parameters(e.g. how many blocks current partition can be moved to get alignment) ?
-----Original Message-----
From: Phillip Susi [mailto:phillsusi <at> gmail.com] On Behalf Of Phil Susi
Sent: Wednesday, December 23, 2015 10:50 PM
To: Ankur Tank; 22180 <at> debbugs.gnu.org
Cc: artfri2 <at> gmail.com
Subject: Re: bug#22180: Partition alignment using parted
On 12/23/2015 8:35 AM, Ankur Tank wrote:
> Hi Phillip
>
> Thank you very much for reply, We are using am335x based (Beaglebone
> Black) custom board with eMMC(4GB). I have couple more questions
> regarding partition alignment. 1. Does partition alignment guarantee
> performance improvement ?
No; it depends entirely on the underlying storage device. Your average hard disk does not care at all about alignment. Hard disks that internally use 4k physical sectors but report 512 byte logical sectors will see a significant performance difference between a partition that is aligned to 4k or not. A raid array performs much better when aligned to a multiple of the stripe size. The performance characteristics of flash vary greatly.
> 2. Is it required for all the partitions to be aligned ? Because we
> are creating four raw partitions at the beginning of the eMMC which
> can't be on 1MB boundary.(because ROM code looks at on 128kb blocks
> only).
No, it is not required. And especially for a mostly read only partition, it really doesn't matter as it is only writes that suffer from improper alignment.
> 3. Is there any better way to get partition aligned. (I know I sound
> silly but I wanted to know if there is any better method ).
I'm not sure what you mean.
> 4.
> parted is not allowing to use ~340 MB of size at the end of the eMMC,
> why so ? Am I missing something ? 5. I don't see verbose option for
> parted in man page, Is there any way to enable it?
I can't read your mind, so you will have to describe what you are actually seeing.
L&T Technology Services Ltd
www.LntTechservices.com<http://www.lnttechservices.com/>
This Email may contain confidential or privileged information for the intended recipient (s). If you are not the intended recipient, please do not use or disseminate the information, notify the sender and delete it from your system.
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Sat, 26 Dec 2015 01:40:02 GMT)
Full text and
rfc822 format available.
Message #20 received at 22180 <at> debbugs.gnu.org (full text, mbox):
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
On 12/24/2015 05:51 AM, Ankur Tank wrote:
> Thank you for replying Phillip, 1. How do we measure performance
> improvement before and after partition alignment ?
hdparm -t or dd between the drive and /dev/null or /dev/zero and the
drive ( for a destructive write test ).
> 2. By above question, I meant, are there any tools/utilities
> available which would suggest partition alignment parameters(e.g.
> how many blocks current partition can be moved to get alignment) ?
Not that I am aware of but it is simply a matter of dividing the
current start position by the desired alignment and seeing if you get
a remainder or fractional result.
parted's align-check command does this, but simply answers with a
yes/no as to whether it is aligned instead of how far off it is.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAEBCgAGBQJWfe+vAAoJEBB5UWFcu6UWcnwH/0Ft7FnlqG3z8WgPyxw/hjtP
B6mf9fyPod/XlExE5YioAMIY/mj7u4AwYBHqvRsr7GqhIxc3PNdT31i1gfwi/khp
sC+nJj39ShBp3C313MyRm9wbUQgMs88qXQAJWjQwRNs/suWyaSodOAzCsJ96aD0D
xiT/9T8s4l3eXSGMeutFCtTRkFh8csOrQTO39Pq6es1uVoZd1fR3VC+HMwfyqt72
CDWhm2EAiAuNBYZ8g0LpD2WLqrh0ZHqui2svLnEZATpc1GsreZyc8jUNDZ9EXFs6
rfQXVXTQqXAHOkQCfF9tD9mLsKUOBN0yDkbl9Zb95gqPRjbqnqEeWq+iWsdSynI=
=GLJ8
-----END PGP SIGNATURE-----
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Sat, 26 Dec 2015 21:45:01 GMT)
Full text and
rfc822 format available.
Message #23 received at submit <at> debbugs.gnu.org (full text, mbox):
On 12/26/2015 12:00 PM, bug-parted-request <at> gnu.org wrote:
>
> On 12/24/2015 05:51 AM, Ankur Tank wrote:
>> Thank you for replying Phillip, 1. How do we measure performance
>> improvement before and after partition alignment ?
>
> hdparm -t or dd between the drive and /dev/null or /dev/zero and the
> drive ( for a destructive write test ).
I don't think you'd get much out of that. The alignment issues are
intimately tied to filesystem design features (and kernel
implementations of same). Thus, I'd suggest copying files between the
partitions you want to test and some other device. Several years ago, I
ran some tests, and eventually updated them on newer disks. The end
result of my tests is written up here:
https://www.ibm.com/developerworks/linux/library/l-linux-on-4kb-sector-disks/
One takeaway from this is that the results vary significantly from one
computer and disk to another, as well as from one filesystem to another.
In most cases, improper-alignment penalties are greatest when writing
small files. When reading files or when writing large files, penalties
are much less severe.
--
Rod Smith
rodsmith <at> rodsbooks.com
Information forwarded
to
bug-parted <at> gnu.org:
bug#22180; Package
parted.
(Sat, 02 Jan 2016 15:28:02 GMT)
Full text and
rfc822 format available.
Message #26 received at 22180 <at> debbugs.gnu.org (full text, mbox):
Hi Phillip,
Finally I could achieve optimal alignment for all partitions except initial 3 partition(which I can't change).
Except initial 3 SPL partitions i have changed minimum partition size to 1MiB and now parted complains about alignment for only initial 3 partitions.
I have updated below Stackexchange post with latest script,
http://unix.stackexchange.com/questions/248939/how-to-achieve-optimal-alignment-for-emmc-partition
However I am not clear with one simple thing, that is,
Why parted behaves different when I give partition size in KiB and MiB?
I have to create partition with size in KiB i have to give end as (end -1) e.g.
parted --script -a optimal /dev/mmcblk0 \
mklabel gpt \
mkpart primary 128KiB 255KiB \
mkpart primary 256KiB 383KiB \
mkpart primary 384KiB 511KiB \
Observe end argument in all 3 partitions they are 256-1, 384-1 and 512-1
However if I have to give size in MiB parted allows me to use exact 1 MiB boundry,
parted --script -a optimal /dev/mmcblk0 \
mklabel gpt \
mkpart primary 128KiB 255KiB \
mkpart primary 256KiB 383KiB \
mkpart primary 384KiB 511KiB \
mkpart primary 1MiB 2MiB \
mkpart primary 2MiB 3MiB \
mkpart primary 3MiB 4MiB \
Observe last three partitions, end of the one partition is the beginning of the second partition in case size mentioned in MiB which is not possible if size is given in KiB.
Why is that, any suggestions/pointers for that ?
Thank you for all help,
Regards,
Ankur
________________________________________
From: Phillip Susi [psusi <at> ubuntu.com]
Sent: Saturday, December 26, 2015 7:08 AM
To: Ankur Tank; 22180 <at> debbugs.gnu.org
Cc: artfri2 <at> gmail.com
Subject: Re: bug#22180: Partition alignment using parted
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
On 12/24/2015 05:51 AM, Ankur Tank wrote:
> Thank you for replying Phillip, 1. How do we measure performance
> improvement before and after partition alignment ?
hdparm -t or dd between the drive and /dev/null or /dev/zero and the
drive ( for a destructive write test ).
> 2. By above question, I meant, are there any tools/utilities
> available which would suggest partition alignment parameters(e.g.
> how many blocks current partition can be moved to get alignment) ?
Not that I am aware of but it is simply a matter of dividing the
current start position by the desired alignment and seeing if you get
a remainder or fractional result.
parted's align-check command does this, but simply answers with a
yes/no as to whether it is aligned instead of how far off it is.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAEBCgAGBQJWfe+vAAoJEBB5UWFcu6UWcnwH/0Ft7FnlqG3z8WgPyxw/hjtP
B6mf9fyPod/XlExE5YioAMIY/mj7u4AwYBHqvRsr7GqhIxc3PNdT31i1gfwi/khp
sC+nJj39ShBp3C313MyRm9wbUQgMs88qXQAJWjQwRNs/suWyaSodOAzCsJ96aD0D
xiT/9T8s4l3eXSGMeutFCtTRkFh8csOrQTO39Pq6es1uVoZd1fR3VC+HMwfyqt72
CDWhm2EAiAuNBYZ8g0LpD2WLqrh0ZHqui2svLnEZATpc1GsreZyc8jUNDZ9EXFs6
rfQXVXTQqXAHOkQCfF9tD9mLsKUOBN0yDkbl9Zb95gqPRjbqnqEeWq+iWsdSynI=
=GLJ8
-----END PGP SIGNATURE-----
L&T Technology Services Ltd
www.LntTechservices.com<http://www.lnttechservices.com/>
This Email may contain confidential or privileged information for the intended recipient (s). If you are not the intended recipient, please do not use or disseminate the information, notify the sender and delete it from your system.
This bug report was last modified 8 years and 76 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.