Discussion:
[Clonezilla-live] Partclone vs Partimage
Jorge Fábregas
2010-04-14 04:39:55 UTC
Permalink
Hello Steven,

I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool. Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.

My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).

Thanks for Clonezillla.

Best regards,
Jorge
Lukas Grässlin
2010-04-14 06:43:20 UTC
Permalink
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.

Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.

Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool. Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Lukas Grässlin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Simply Linux.

Geschäftsführer: Boris Nalbach
AG München HRB 158898 * Ust.-IdNr: DE 814464942
Steven Shiau
2010-04-14 08:33:36 UTC
Permalink
Yes, Thomas Tsai is working on the improvement of partclone.

Lukas,
Please send us gprof results you have.
Thanks.

Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool. Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Lukas Grässlin
2010-04-14 09:34:35 UTC
Permalink
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool. Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Lukas GrÀsslin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Simply Linux.

GeschÀftsfÌhrer: Boris Nalbach
AG MÃŒnchen HRB 158898 * Ust.-IdNr: DE 814464942
Jorge Fábregas
2010-04-14 13:02:58 UTC
Permalink
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Thanks Steven. And the "Calculating Bitmap" step...What is this step actually
doing?

Best regards,
Jorge
Steven Shiau
2010-04-14 13:44:55 UTC
Permalink
Post by Jorge Fábregas
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Thanks Steven. And the "Calculating Bitmap" step...What is this step actually
doing?
Partclone need to know where the used blocks are in the partition, so
first it searches them.
BTW, which version of Clonezilla live are you using? Did you try 1.2.5-1?

Steven.
Post by Jorge Fábregas
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Jorge Fábregas
2010-04-14 23:38:08 UTC
Permalink
Post by Steven Shiau
Partclone need to know where the used blocks are in the partition, so
first it searches them.
Ahh got it. Thanks for the explanation.
Post by Steven Shiau
BTW, which version of Clonezilla live are you using? Did you try 1.2.5-1?
I tried the latest stable (1.2.4-28). Hmm ok I'll give 1.2.5-1 a try and will
report back.

Thanks!

Best regards,
Jorge
Jorge Fábregas
2010-04-15 23:22:07 UTC
Permalink
Post by Steven Shiau
Did you try 1.2.5-1?
Wow! I tried 1.2.5-2 today and it was about 200% faster! Here are the test
results:

For both tests, I used the following:

- partition image backup to SSH server thru Gigabit link
- Filessytem: ext3
- Device size: 58 GB
- Space in use: 3.5 GB
- Block size: 4096 bytes
- size in MB to split partition image: 1000000 (to avoid split)
- Compression on image: none
- the rest of settings were default ones

For the above scenario, the times were:

a) Using clonezilla-live-1.2.4-28-686.iso, it took 8 minutes and 5 seconds

b) Using clonezilla-live-1.2.5-2.686.iso, it took 2 minutes and 31 seconds

The elapsed time above is just after hitting "Are you sure ..." where you
press "y" until the end of the image creation + hw gathering info output etc,
specifically "Press Enter to continue...".

Impressive really. Excellent job guys!

So when does "testing" becomes "stable". Is it based on some time or is it if
it's been there for some time without many people reporting any issues?

All the best,
Jorge
Steven Shiau
2010-04-17 04:01:33 UTC
Permalink
Hi Jorge,
Post by Jorge Fábregas
Post by Steven Shiau
Did you try 1.2.5-1?
Wow! I tried 1.2.5-2 today and it was about 200% faster! Here are the test
- partition image backup to SSH server thru Gigabit link
- Filessytem: ext3
- Device size: 58 GB
- Space in use: 3.5 GB
- Block size: 4096 bytes
- size in MB to split partition image: 1000000 (to avoid split)
- Compression on image: none
- the rest of settings were default ones
a) Using clonezilla-live-1.2.4-28-686.iso, it took 8 minutes and 5 seconds
b) Using clonezilla-live-1.2.5-2.686.iso, it took 2 minutes and 31 seconds
The elapsed time above is just after hitting "Are you sure ..." where you
press "y" until the end of the image creation + hw gathering info output etc,
specifically "Press Enter to continue...".
Impressive really. Excellent job guys!
Cool. Thanks for confirming that. Actually partclone still has room to
be improved. Thomas Tsai will try to make it recently.
Post by Jorge Fábregas
So when does "testing" becomes "stable". Is it based on some time or is it if
it's been there for some time without many people reporting any issues?
It depends. Normally we will release a stable one in about 2 months. If
people report any issue, definitely we will do our best to fix it.
Actually if the version you are using did the job for you, stick with it.

Steven.
Post by Jorge Fábregas
All the best,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Steven Shiau
2010-04-14 13:42:21 UTC
Permalink
These are results I got from testing today. (Cant tell you why gprof now
lists all the strcmp, strlen etc functions. In my first test it only
showed me the functions from partclone itself, however)
(normally the crc32 function would be on the top but i commented its
functionallity out for testing)
0.00 0.73 0.00 690183 0.00 0.00 log_mesg
0.00 0.73 0.00 131073 0.00 0.00 Ncurses_progress_update
0.00 0.73 0.00 131073 0.00 0.00 calculate_speed
0.00 0.73 0.00 131073 0.00 0.00 update_pui
2.94 0.55 0.02 1036528 0.00 0.00 log_mesg
0.00 0.68 0.00 262146 0.00 0.00 progress_update
0.00 0.68 0.00 262146 0.00 0.00 update_pui
As you can see these functions are called very often which affects (at
least on my machines) in very high cpu load.
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.

Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool. Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Lukas Grässlin
2010-04-14 14:04:10 UTC
Permalink
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.

If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Lukas Grässlin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Simply Linux.

Geschäftsführer: Boris Nalbach
AG München HRB 158898 * Ust.-IdNr: DE 814464942
Steven Shiau
2010-04-14 14:14:52 UTC
Permalink
Post by Lukas Grässlin
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thomas,
Please make sure "the update_pui method which also runs calculate_speed
anyway is run". If it's true, I believe this can be improved...
Thanks.

Steven.
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Steven Shiau
2010-04-14 14:15:16 UTC
Permalink
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Lukas Grässlin
2010-04-14 14:53:09 UTC
Permalink
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.

You can see it e.g. here in restore.c :

/// start restore image file to partition
for( block_id = 0; block_id < image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for


I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like
calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Lukas Grässlin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Simply Linux.

Geschäftsführer: Boris Nalbach
AG München HRB 158898 * Ust.-IdNr: DE 814464942
Steven Shiau
2010-04-14 21:16:59 UTC
Permalink
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id < image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.

Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like
calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual creation of the
image). I'm a bit confused about these 2 steps as soon as partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Lukas Grässlin
2010-04-15 06:55:10 UTC
Permalink
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id < image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.
I got about 20-39 MB/s.
Here is the patch. There are some other changes within:

* I completeley disabled the update_pui on the server side, because
clonezilla anyway doesn't show the output there.

* I changed the display of the speed from */min to */s beaucse I thoght
it's better readable for the most people.

* I implemented that crc32-checks can be disabled. Not an good idea, but
I wanted it for testing and its disabled by default.

Sorry, I had no time to rip off these other changes but it should be ok
anyway I hope.

Regards,
Lukas
Post by Steven Shiau
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like
calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual
creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Lukas GrÀsslin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Simply Linux.

GeschÀftsfÌhrer: Boris Nalbach
AG MÃŒnchen HRB 158898 * Ust.-IdNr: DE 814464942
Steven Shiau
2010-04-17 03:10:49 UTC
Permalink
Lukas,
Thanks. Thomas Tsai will revise this and try to use them.
In the future, we will have more options in partclone to disable some
jobs, e.g. "-q, --quiet" to skip all the rate messages update.
Thanks.

Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id < image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.
I got about 20-39 MB/s.
* I completeley disabled the update_pui on the server side, because
clonezilla anyway doesn't show the output there.
* I changed the display of the speed from */min to */s beaucse I thoght
it's better readable for the most people.
* I implemented that crc32-checks can be disabled. Not an good idea, but
I wanted it for testing and its disabled by default.
Sorry, I had no time to rip off these other changes but it should be ok
anyway I hope.
Regards,
Lukas
Post by Steven Shiau
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after that it
started again another progress indicator (I guess the actual
creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Lukas Grässlin
2010-10-04 11:56:44 UTC
Permalink
Any news about that?
Post by Steven Shiau
Lukas,
Thanks. Thomas Tsai will revise this and try to use them.
In the future, we will have more options in partclone to disable
some jobs, e.g. "-q, --quiet" to skip all the rate messages update.
Thanks.
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id < image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.
I got about 20-39 MB/s.
* I completeley disabled the update_pui on the server side, because
clonezilla anyway doesn't show the output there.
* I changed the display of the speed from */min to */s beaucse I thoght
it's better readable for the most people.
* I implemented that crc32-checks can be disabled. Not an good idea, but
I wanted it for testing and its disabled by default.
Sorry, I had no time to rip off these other changes but it should be ok
anyway I hope.
Regards,
Lukas
Post by Steven Shiau
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use this.
My observation was that partclone takes some more time. When it
started
saving the partition, I wasn't sure what it was doing. There was some
progress indicator (percentage) (Generating bitmap..) and I after
that it
started again another progress indicator (I guess the actual
creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
--
Lukas Grässlin

Collax GmbH . Basler Str. 115a . 79115 Freiburg . Germany

p: +49 (0) 89-990 157-23

Collax - Flexible IT.

Geschäftsführer: Bernd Bönte, Boris Nalbach
Amtsgericht München, HRB 173695
USt-ID: DE270819312
Steven Shiau
2010-10-05 08:07:34 UTC
Permalink
Hi Lukas,
Thanks for reminding us this.
We will add this in the next release.

Steven.
Post by Lukas Grässlin
Any news about that?
Post by Steven Shiau
Lukas,
Thanks. Thomas Tsai will revise this and try to use them.
In the future, we will have more options in partclone to disable
some jobs, e.g. "-q, --quiet" to skip all the rate messages update.
Thanks.
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id< image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.
I got about 20-39 MB/s.
* I completeley disabled the update_pui on the server side, because
clonezilla anyway doesn't show the output there.
* I changed the display of the speed from */min to */s beaucse I thoght
it's better readable for the most people.
* I implemented that crc32-checks can be disabled. Not an good idea, but
I wanted it for testing and its disabled by default.
Sorry, I had no time to rip off these other changes but it should be ok
anyway I hope.
Regards,
Lukas
Post by Steven Shiau
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use
this.
My observation was that partclone takes some more time. When it
started
saving the partition, I wasn't sure what it was doing. There was
some
progress indicator (percentage) (Generating bitmap..) and I after
that it
started again another progress indicator (I guess the actual
creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau<steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
--
Steven Shiau <steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan.
http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Steven Shiau
2010-10-08 06:57:19 UTC
Permalink
Hi Lukas,
The partclone 0.2.16 has an option now:
-q, --quiet Disable progress message

Steven.
Post by Lukas Grässlin
Any news about that?
Post by Steven Shiau
Lukas,
Thanks. Thomas Tsai will revise this and try to use them.
In the future, we will have more options in partclone to disable
some jobs, e.g. "-q, --quiet" to skip all the rate messages update.
Thanks.
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
Post by Steven Shiau
There is an option "-f" of partclone which you might be interested to
give it try.
You can tune it to see if any difference.
As I understood the code, -f only affects the time the gui itself would
be refreshed, but the update_pui method which also runs calculate_speed
anyway is run.
Thanks. I will check with Thomas,
No problem.
/// start restore image file to partition
for( block_id = 0; block_id< image_hdr.totalblock; block_id++ ){
/* doing things, copying the blocks */
update_pui(&prog, copied, done);
} // end of for
I did a patch for myself within I told it just call update_pui once for
5000 blocks. Don't know if its a good idea, but it worked ;)
Good, and the performance is?
Could you please also send us the patch file?
Thanks.
I got about 20-39 MB/s.
* I completeley disabled the update_pui on the server side, because
clonezilla anyway doesn't show the output there.
* I changed the display of the speed from */min to */s beaucse I thoght
it's better readable for the most people.
* I implemented that crc32-checks can be disabled. Not an good idea, but
I wanted it for testing and its disabled by default.
Sorry, I had no time to rip off these other changes but it should be ok
anyway I hope.
Regards,
Lukas
Post by Steven Shiau
Steven.
Post by Lukas Grässlin
Post by Steven Shiau
Post by Lukas Grässlin
If I've time I'll try the older clonezilla live, too.
Post by Steven Shiau
BTW, maybe you can also give Clonezilla live 1.2.2-14 a try? It's
partclone is older, and we might have a regression somewhere...
Please let us know the results.
Thanks.
Regards,
Steven.
Post by Steven Shiau
Yes, Thomas Tsai is working on the improvement of partclone.
Lukas,
Please send us gprof results you have.
Thanks.
Steven.
Post by Lukas Grässlin
Regarding the perfmance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
Regards,
Lukas
Post by Jorge Fábregas
Hello Steven,
I've been using Clonezilla happily for more than 2 years and I
always used the
custom options and specifically partimage as the cloning tool.
Since partimage
seems to be a dead project (and no support for ext4) and
considering that
Clonezilla uses partclone as the default option I decided to use
this.
My observation was that partclone takes some more time. When it
started
saving the partition, I wasn't sure what it was doing. There was
some
progress indicator (percentage) (Generating bitmap..) and I after
that it
started again another progress indicator (I guess the actual
creation of the
image). I'm a bit confused about these 2 steps as soon as
partclone starts.
Any tip will be appreciated. (just curious).
Thanks for Clonezillla.
Best regards,
Jorge
------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Clonezilla-live mailing list
https://lists.sourceforge.net/lists/listinfo/clonezilla-live
--
Steven Shiau<steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
--
Steven Shiau<steven _at_ nchc org tw> <steven _at_ stevenshiau org>
National Center for High-performance Computing, Taiwan. http://www.nchc.org.tw
Public Key Server PGP Key ID: 1024D/9762755A
Fingerprint: A2A1 08B7 C22C 3D06 34DB F4BC 08B3 E3D7 9762 755A
Jorge Fábregas
2010-04-14 12:49:06 UTC
Permalink
Regarding the performance of partclone: Look at the "on-the-fly
performance" Thread in this mailing list.
Thanks for the tip Lukas. Indeed, an interesting thread. Exactly what I wanted
to know since I almost experienced the same throughput as you did (compared to
partimage).
Partclone does some odd things which slow down the speed. You can
improve this by saying not to use the gui. (I think this is in the
expert options). But it stills does stupid things, like calculating the
speed too often which resultes in high cpu load.
I'll try the nogui option next time then. Thanks!

Best regards,
Jorge
Loading...