Il 04/12/2016 06:25, Robert Munteanu ha scritto:

Hi Marco,

...

Dears Tumbleweed users, gurus and hackers,

I would like to get some suggestions by more skilled people in order to improve my low IO disk performance because sometimes by using iotop I can see over 99% of disk activity especially when running VirtualBox and packagekit is updating the system.

HDD activity led is always on during high IO activity and system sometimes get unresponsive or in the better case very very slow.

I tried to change one ext4 parameter from [cfq] to [noop] without any noticeable improvements.

sudo echo noop > /sys/block/sda/queue/scheduler ISTR that noop would be beneficial for SSDs, but not for HDDs. Other

On Fri, Dec 2, 2016 at 2:05 PM, Marco Calistri wrote: than that, I suggest that you use general Linux tuning guides available online, I don't think there are any specific openSUSE tweaks.

Best,

Robert

...

Would like to know if I can try some other tuning.

My HDD is an ATA 465,8G SAMSUNG HM501II ext4

Many thanks and regards!

Cheers,

-- Marco Calistri Opensuse Tumbleweed 64 bit - Kernel 4.7.0-2-default Gnome 3.20 Intel® Core™ i5-2410M CPU @ 2.30GHz × 4 - Intel® Sandybridge Mobile

Thanks for your feedback Robert! I rolled back to [cfq] now but want to try the parameter vm.swappiness which as default is =60 and set it to =10: cat /etc/sysctl.d/99-sysctl.conf vm.swappiness=10 Cheers, Marco N�����r��y隊Z)z{.���r�+�맲��r��z�^�ˬz��N�(�֜��^� ޭ隊Z)z{.���r�+��0�����Ǩ�

On Sun, 2016-12-04 at 13:49 +0000, Marco Calistri wrote:

I rolled back to [cfq] now but want to try the parameter vm.swappiness which as default is =60 and set it to =10:

cat /etc/sysctl.d/99-sysctl.conf

vm.swappiness=10

Rather than playing with tuning parameters, I'd suggest to proceed more systematically. Run some benchmarks to find out the "raw" performance of your disk for sequential and random IO. Do all this with the "noop" scheduler. I usually start out with a plain "dd" sequential read from the block device with 1M block size. If you have an unused partition, try sequential writes as well. Compare that to the device specs (or possibly with measurements you find on the net). This will give you an idea what to expect from the HW, and whether the HW is operating in the expected speed range. Next, run some simple benchmarks like bonnie++ on your file system. Be sure not to measure the page cache (use direct IO). Try different IO schedulers to see if it makes a difference. Finally, under your problematic load (virtualbox + packagekit), run iostat, vmstat or something similar and compare the disk throughput numbers you obtain with those from the previous steps. If the numbers (roughly) match the values to be expected from the HW, you'll either have to look for faster storage HW, or you need to distribute your IO load better. Especially if you have lots of RAM a lot of writes, you may want to look into the vm.dirty* sysctl settings. My personal rule of thumb is "set dirty_bytes to the number of bytes that the IO HW can write in ~1 second, and dirty_background_bytes to 10x the previous value". Regards Martin

Cheers,

Marco Nry隊Z)z{.r+맲rz^ˬzN(֜^ ޭ隊Z)z{.r+0Ǩ

-- Dr. Martin Wilck , Tel. +49 (0)911 74053 2107 SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

On 2016-12-07 11:53, Martin Wilck wrote:

Rather than playing with tuning parameters, I'd suggest to proceed more systematically. Run some benchmarks to find out the "raw" performance of your disk for sequential and random IO. Do all this with the "noop" scheduler. I usually start out with a plain "dd" sequential read from the block device with 1M block size. If you have an unused partition, try sequential writes as well. Compare that to the device specs (or possibly with measurements you find on the net). This will give you an idea what to expect from the HW, and whether the HW is operating in the expected speed range.

You can measure the raw read speed with hdparm -Tt /dev/DEVICE -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" at Telcontar)

Il 09/12/2016 00:49, Marco Calistri ha scritto:

marco@linux-turion64:~> sudo hdparm -Tt /dev/sda [sudo] password di root:

/dev/sda: Timing cached reads: 9168 MB in 2.00 seconds = 4585.57 MB/sec Timing buffered disk reads: 232 MB in 3.02 seconds = 76.71 MB/sec

HD Model ? It seems slow.. Here with an old Seagate Barracuda: /dev/sda: Timing cached reads: 13828 MB in 2.00 seconds = 6920.66 MB/sec Timing buffered disk reads: 308 MB in 3.00 seconds = 102.53 MB/sec Bye, Daniele. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Il 08/12/2016 22:09, Marco Calistri ha scritto:

Il 08/12/2016 22:03, Daniele ha scritto:

...

Il 09/12/2016 00:49, Marco Calistri ha scritto:

...

marco@linux-turion64:~> sudo hdparm -Tt /dev/sda [sudo] password di root:

/dev/sda: Timing cached reads: 9168 MB in 2.00 seconds = 4585.57 MB/sec Timing buffered disk reads: 232 MB in 3.02 seconds = 76.71 MB/sec HD Model ? It seems slow..

Here with an old Seagate Barracuda: /dev/sda: Timing cached reads: 13828 MB in 2.00 seconds = 6920.66 MB/sec Timing buffered disk reads: 308 MB in 3.00 seconds = 102.53 MB/sec

Bye, Daniele. Hi Daniele, It is a SAMSUNG 500G, SAMSUNG HM501II ATA.

Cheers,

Forgot to say that I'm using ext4 here! -- Marco Calistri Opensuse Tumbleweed 64 bit Intel® Core™ i5-2410M CPU @ 2.30GHz × 4 Intel® Sandybridge Mobile N�����r��y隊Z)z{.���r�+�맲��r��z�^�ˬz��N�(�֜��^� ޭ隊Z)z{.���r�+��0�����Ǩ�

------------------------------------------------------------------------ *From:* Daniele *Sent:* 08/12/2016 22:18:52 (-0200 UTC) *To:* Opensuse-factory *Subject:* Re: [opensuse-factory] Improving IO with Sata HDD

Il 09/12/2016 01:13, Marco Calistri ha scritto:

...

...

It is a SAMSUNG 500G, SAMSUNG HM501II ATA.

Cheers,

Forgot to say that I'm using ext4 here!

HDD is slow, nothing to do.. It's time to move on SSD or Hybrid if you want better performance.

Daniele. Got it, Thanks! N�����r��y隊Z)z{.���r�+�맲��r��z�^�ˬz��N�(�֜��^� ޭ隊Z)z{.���r�+��0�����Ǩ�

On 9 December 2016 at 09:46, Carlos E. R. wrote:

How many RPM?

Maybe it is a 5200 RPM model. That speed would be typical of such.

Searching it up online, it seems to be a "5400 RPM 8MB Cache SATA 3.0Gb/s 2.5" Internal Notebook Hard Drive", which would (at least partially) explain things. - Karl Cheng (Qantas94Heavy) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2016-12-09 13:05, Marco Calistri wrote:

*From:* Carlos E. R.

...

How many RPM?

Maybe it is a 5200 RPM model. That speed would be typical of such.

Yes of course is the slower speed, I would be very surprised it being the 7200 RPM model. Anyhow with bfq scheduler is accettable this is a notebook and not a multi-core 24x8 server ;-)

Well, your disk has the normal speed for its class. You can use a disk with a faster rotating speed, which will transfer data faster, and also use more battery. Or switch to a flash disk (watch the sustained write speed), which is faster but has a shorter life and is more expensive for the same size. Uses less battery, I believe. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlhKoA8ACgkQja8UbcUWM1zhtwD/S6N+ksxumlUf/Pvoo+hdqgUJ UxgE4WhM6dDTkzcnupIA/0dz4UOm5N3JL5Wr2z+uGW9EZAPSAYP8nABaIrHwO+UH =5J+I -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2016-12-09 13:21, Jan Engelhardt wrote:

On Friday 2016-12-09 13:05, Marco Calistri wrote:

...

...

How many RPM? Maybe it is a 5200 RPM model. That speed would be typical of such.

Yes of course is the slower speed, I would be very surprised it being the 7200 RPM model.

Rotational rate however has more impact on random reads (seeks) than transfer speed for streaming (linear) reads.

On both. hdparm -tT goes up to 100 MB/s at typically. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlhKpYUACgkQja8UbcUWM1ztowD9FyH23s+wTo21EXgcrPMJWR4L vnNZ7HuOLh5U+8uJHsMA/3Ob/6kikzI8O5gO/KlHncGFAGiXWZpU7cWC6eCF+pNT =Bfyw -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2016-12-09 00:49, Marco Calistri wrote:

Carlos, here you can see the result of the command you suggested:

marco@linux-turion64:~> sudo hdparm -Tt /dev/sda [sudo] password di root:

/dev/sda: Timing cached reads: 9168 MB in 2.00 seconds = 4585.57 MB/sec Timing buffered disk reads: 232 MB in 3.02 seconds = 76.71 MB/sec

An interesting exercise is to partition an empty disk in, say, 20 equally sized partitions, and measure the speed on each one. And do it for different filesystems, too. You probably will find out that the disk is faster about 1/3 of the way. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlhKCuUACgkQja8UbcUWM1yVXgD+ILRbp33EXKCmHEt8JE1now+q wJ6GNueGZbQ9I/bs5YMA/jLGMJPF64B+d8qnNSKDCXMVUCR5A/X12gg6oG1LdzD2 =MjAb -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 2016-12-09 21:13, Martin Wilck wrote:

On Fri, 2016-12-09 at 02:37 +0100, Carlos E. R. wrote:

...

An interesting exercise is to partition an empty disk in, say, 20 equally sized partitions, and measure the speed on each one. And do it for different filesystems, too.

You probably will find out that the disk is faster about 1/3 of the way.

True for sequential IO, for obvious reasons (more data on a single track in the outer part of the disk). For random IO, you'll have an average wait time of a half rotation, plus head movement, and both is independent on the length of the track.

The curious thing is that outer part of the disk is often slower than some way into the disk. - -- Cheers / Saludos, Carlos E. R. (from 13.1 x86_64 "Bottle" (Minas Tirith)) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlhLH/IACgkQja8UbcUWM1zlTgD7BSKnLHs3gk0Ucg6MNXYsuqb2 F/iGgrnyDFEF2ijcwIsA/2o5epJ1xP+n9tIMVRVLXBny32VXoEXr1C8dTD4EKHMV =HFtn -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Il 04/12/2016 07:06, Daniele ha scritto:

Il 02/12/2016 13:05, Marco Calistri ha scritto:

...

Dears Tumbleweed users, gurus and hackers,

I would like to get some suggestions by more skilled people in order to improve my low IO disk performance because sometimes by using iotop I can see over 99% of disk activity especially when running VirtualBox and packagekit is updating the system.

HDD activity led is always on during high IO activity and system sometimes get unresponsive or in the better case very very slow.

I tried to change one ext4 parameter from [cfq] to [noop] without any noticeable improvements.

sudo echo noop > /sys/block/sda/queue/scheduler

Would like to know if I can try some other tuning.

My HDD is an ATA 465,8G SAMSUNG HM501II ext4

Many thanks and regards!

Cheers,

I don't know if is a valid advise for TW but on os 13.2 I am using the BFQ scheduler.

Bye, Daniele.

Hi Daniele, BFQ is not available on m TW so can't give it a try. marco@linux-turion64:~> cat /sys/block/sda/queue/scheduler noop deadline [cfq] Thanks and regards, -- Marco Calistri Opensuse Tumbleweed 64 bit Intel® Core™ i5-2410M CPU @ 2.30GHz × 4 Intel® Sandybridge Mobile

Il 04/12/2016 12:06, Daniele ha scritto:

Il 04/12/2016 14:53, Marco Calistri ha scritto:

...

Hi Daniele,

BFQ is not available on m TW so can't give it a try. marco@linux-turion64:~> cat /sys/block/sda/queue/scheduler noop deadline [cfq]

Thanks and regards,

Yes, not available (by default) on 13.2 too. Search on software.opensuse.org !

Bye, Daniele.

Ok! Thanks and regards, -- Marco Calistri Opensuse Tumbleweed 64 bit Intel® Core™ i5-2410M CPU @ 2.30GHz × 4 Intel® Sandybridge Mobile

Il 04/12/2016 16:09, Anton Aylward ha scritto:

On 12/04/2016 04:06 AM, Daniele wrote:

...

...

I don't know if is a valid advise for TW but on os 13.2 I am using the BFQ scheduler.

Do you mean CFQ? http://www.phoronix.com/scan.php?page=news_item&px=cfq-io-scheduler-iops-linux-4.2

Nope, BFQ ! https://www.phoronix.com/scan.php?page=news_item&px=BFQ-V8-Replacing-CFQ https://software.opensuse.org/package/bfq Bye, Daniele. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

2016-12-04 13:05 GMT-03:00 Marco Calistri :

Il 04/12/2016 13:15, Daniele ha scritto:

...

Il 04/12/2016 16:09, Anton Aylward ha scritto:

...

On 12/04/2016 04:06 AM, Daniele wrote:

...

...

I don't know if is a valid advise for TW but on os 13.2 I am using the BFQ scheduler.

Do you mean CFQ? http://www.phoronix.com/scan.php?page=news_item&px=cfq-io-scheduler-iops-linux-4.2

Nope, BFQ !

https://www.phoronix.com/scan.php?page=news_item&px=BFQ-V8-Replacing-CFQ

https://software.opensuse.org/package/bfq

Bye, Daniele. Daniele, I installed available bfq as kernel module but so far I'm not noticing any impressive performance. I started three consoles, one Virtualbox guest (windowx XP) Thunderbird and Firefox and I observed a long period of HDD high spinning and a bit of system freezing. I still have the parameter vm.swappiness=10 I don't know if I have to set it as per default. Anyhow I will let bfq scheduler activated for a week or two to see if it is better or not for my installation. marco@linux-turion64:~> cat /sys/block/sda/queue/scheduler [bfq] noop deadline cfq

Hi! You forgot to tell us how much memory has your system. Because with less memory system You have higher IO on the hard disk. Regards, Juan -- USA LINUX OPENSUSE QUE ES SOFTWARE LIBRE, NO NECESITAS PIRATEAR NADA Y NI TE VAS A PREOCUPAR MAS POR LOS VIRUS Y SPYWARES: http://www.opensuse.org/es/ Puedes visitar mi blog en: http://jerbes.blogspot.com.ar/ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org