Re: [Soekris] Poor I/O performance
Am 12.03.2014 10:37, schrieb Bob Bishop: > If your software doesn't support TRIM then SSD write performance (and read, > to a lesser extent) will start to suck once the disk has had a certain amount > of use. Right! During initial installation i made an image. First i filled the empty space with zeros: for i in $(seq 100) ; do dd if=/dev/zero of=/$i.dd bs=1M count=1024 done The idea was to not have empty space with junk data in it, because that would not compress very good. Then i made the image: dd if=/dev/sda bs=10M | gzip --fast | ssh -c arcfour kkr...@nb8810.qitsloc.ratq.local "cat > /home/kkruse/img.dd.gz" The result was a 900 mb file. The command to write it back: ssh -c arcfour kkr...@nb8810.qitsloc.ratq.local "cat /home/kkruse/img.dd.gz" | gzip --decompress | dd of=/dev/sda This is a SSD nightmare, i guess. ;) I just used fstrim to tell the SSD that almost all of it is empty space, for the future i added the discard mount option. The write performance went from 20 MB/s to 30 MB/s. Not perfect, but at least a bit better. MFG, Karsten Kruse -- Karsten Kruse System Engineer Unix/Linux Operation Center Infrastructure Telefon: +49 (0) 2102 - 852 - 4335 (- 202 Fax) karsten.kr...@qits.de http://www.qits.de/ QITS GmbH, Halskestraße 11-15 - 40880 Ratingen __ Geschäftsführer: Martin Giebel, Bernd Salz Handelsregister: AG Düsseldorf HRB 44009 Ust.ID.-Nr.: DE 196843119 - Steuer-Nr.: 147/5761/047 0x35DBCAEC.asc Description: application/pgp-keys signature.asc Description: OpenPGP digital signature ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] Poor I/O performance
On Wed 12 Mar 2014 09:37:19 AM Bob Bishop wrote: > Hi, > > If your software doesn't support TRIM then SSD write performance (and read, > to a lesser extent) will start to suck once the disk has had a certain > amount of use. Don't know if this applies in your case. SSDs are rather decent about managing themselves these days. But I got the same meager write performance with my 6501-50 on a fresh Crucial M4 mSATA SSD that should be capable of 240MB/s. I'm tentatively blaming the sata controller built into the chipset, or possibly the SoC firmware not initializing things correctly. Or maybe, the platform as it is wired up, just doesn't have the bandwidth for decent speeds off the mSATA ports? > On 11 Mar 2014, at 10:32, Karsten Kruse wrote: > > Ahoi, > > > > i have a bunch of net6501-70 for evaluation and they show poor I/O > > performance. > > > > I use CentOS 6.5 with kernel 2.6.32-431.5.1.el6.x86_64 and this is my > > disk: > > > > > > [root@sonde04 ~]# hdparm -i /dev/sda > > > > /dev/sda: > > > > Model=Samsung SSD 840 EVO 120GB, FwRev=EXT0BB0Q, SerialNo=S1D5NSADB40689E > > Config={ Fixed } > > RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0 > > BuffType=unknown, BuffSize=unknown, MaxMultSect=16, MultSect=16 > > CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=234441648 > > IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120} > > PIO modes: pio0 pio1 pio2 pio3 pio4 > > DMA modes: mdma0 mdma1 mdma2 > > UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 > > AdvancedPM=no WriteCache=enabled > > Drive conforms to: unknown: ATA/ATAPI-2,3,4,5,6,7 > > > > * signifies the current active mode > > > > > > Looks good to me, UDMA6 is used. This doesn't look as good: > > > > > > [root@sonde04 ~]# time dd if=/dev/sda of=/dev/null bs=1M count=1024 > > 1024+0 Datensätze ein > > 1024+0 Datensätze aus > > 1073741824 Bytes (1,1 GB) kopiert, 14,0558 s, 76,4 MB/s > > > > > > So this is read, this is what write looks like: > > > > > > [root@sonde04 ~]# sync ; time { dd if=/dev/zero of=/tmp/test.dd bs=1M > > count=1024 ; sync ; } > > 1024+0 Datensätze ein > > 1024+0 Datensätze aus > > 1073741824 Bytes (1,1 GB) kopiert, 47,0048 s, 22,8 MB/s > > > > real0m50.428s > > user0m0.022s > > sys 0m7.021s > > > > > > This is 1 GB (power of 2) written in 50 seconds, so that amounts to > > about 20,5 MB/s. Would anyone be so kind and run the commands himself, > > maybe with Debian or something with a more recent kernel. > > > > Also i would appreciate any tips on how to get a bit more write > > performance. We would like to monitor traffic with it, using NtopNG and > > tcpdump. When we dump traffic, we don't write payloads, just headers, so > > we don't need more than maybe 40 MB/s guess. > > > > > > Have a nice day, > > > > Karsten Kruse > > -- > Bob Bishop > r...@gid.co.uk > > > > > ___ > Soekris-tech mailing list > Soekris-tech@lists.soekris.com > http://lists.soekris.com/mailman/listinfo/soekris-tech -- Thomas Fjellstrom tho...@fjellstrom.ca ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] Poor I/O performance
Hi, If your software doesn't support TRIM then SSD write performance (and read, to a lesser extent) will start to suck once the disk has had a certain amount of use. Don't know if this applies in your case. On 11 Mar 2014, at 10:32, Karsten Kruse wrote: > > Ahoi, > > i have a bunch of net6501-70 for evaluation and they show poor I/O > performance. > > I use CentOS 6.5 with kernel 2.6.32-431.5.1.el6.x86_64 and this is my disk: > > > [root@sonde04 ~]# hdparm -i /dev/sda > > /dev/sda: > > Model=Samsung SSD 840 EVO 120GB, FwRev=EXT0BB0Q, SerialNo=S1D5NSADB40689E > Config={ Fixed } > RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0 > BuffType=unknown, BuffSize=unknown, MaxMultSect=16, MultSect=16 > CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=234441648 > IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120} > PIO modes: pio0 pio1 pio2 pio3 pio4 > DMA modes: mdma0 mdma1 mdma2 > UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 > AdvancedPM=no WriteCache=enabled > Drive conforms to: unknown: ATA/ATAPI-2,3,4,5,6,7 > > * signifies the current active mode > > > Looks good to me, UDMA6 is used. This doesn't look as good: > > > [root@sonde04 ~]# time dd if=/dev/sda of=/dev/null bs=1M count=1024 > 1024+0 Datensätze ein > 1024+0 Datensätze aus > 1073741824 Bytes (1,1 GB) kopiert, 14,0558 s, 76,4 MB/s > > > So this is read, this is what write looks like: > > > [root@sonde04 ~]# sync ; time { dd if=/dev/zero of=/tmp/test.dd bs=1M > count=1024 ; sync ; } > 1024+0 Datensätze ein > 1024+0 Datensätze aus > 1073741824 Bytes (1,1 GB) kopiert, 47,0048 s, 22,8 MB/s > > real0m50.428s > user0m0.022s > sys 0m7.021s > > > This is 1 GB (power of 2) written in 50 seconds, so that amounts to > about 20,5 MB/s. Would anyone be so kind and run the commands himself, > maybe with Debian or something with a more recent kernel. > > Also i would appreciate any tips on how to get a bit more write > performance. We would like to monitor traffic with it, using NtopNG and > tcpdump. When we dump traffic, we don't write payloads, just headers, so > we don't need more than maybe 40 MB/s guess. > > > Have a nice day, > > Karsten Kruse > > > -- > Karsten Kruse > System Engineer Unix/Linux > Operation Center Infrastructure > > Telefon: +49 (0) 2102 - 852 - 4335 (- 202 Fax) > karsten.kr...@qits.de > http://www.qits.de/ > > QITS GmbH, Halskestraße 11-15 - 40880 Ratingen > __ > > Geschäftsführer: Martin Giebel, Bernd Salz > Handelsregister: AG Düsseldorf HRB 44009 > Ust.ID.-Nr.: DE 196843119 - Steuer-Nr.: 147/5761/047 > <0x35DBCAEC.asc>___ > Soekris-tech mailing list > Soekris-tech@lists.soekris.com > http://lists.soekris.com/mailman/listinfo/soekris-tech -- Bob Bishop r...@gid.co.uk ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] Poor I/O performance
On Tue 11 Mar 2014 11:32:37 AM Karsten Kruse wrote: > Ahoi, > > i have a bunch of net6501-70 for evaluation and they show poor I/O > performance. > > I use CentOS 6.5 with kernel 2.6.32-431.5.1.el6.x86_64 and this is my disk: > Linux firewall 3.12.0-rc3+ #3 SMP Mon Oct 21 17:15:52 MDT 2013 x86_64 GNU/Linux I get about the same as you. My read speeds are higher, around 130MB/s, but testing write speeds to my Crucial m4 gets me 20MB/s. I thought that was more or less normal for the mSATA slots. Maybe I'm wrong though. I certainly would expect that m4 card to get much higher than 20MB/s write on its own, so it likely isn't the card that's slow. Specs say 260MB/s. > [root@sonde04 ~]# hdparm -i /dev/sda > > /dev/sda: > > Model=Samsung SSD 840 EVO 120GB, FwRev=EXT0BB0Q, SerialNo=S1D5NSADB40689E > Config={ Fixed } > RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0 > BuffType=unknown, BuffSize=unknown, MaxMultSect=16, MultSect=16 > CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=234441648 > IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120} > PIO modes: pio0 pio1 pio2 pio3 pio4 > DMA modes: mdma0 mdma1 mdma2 > UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 > AdvancedPM=no WriteCache=enabled > Drive conforms to: unknown: ATA/ATAPI-2,3,4,5,6,7 > > * signifies the current active mode > > > Looks good to me, UDMA6 is used. This doesn't look as good: > > > [root@sonde04 ~]# time dd if=/dev/sda of=/dev/null bs=1M count=1024 > 1024+0 Datensätze ein > 1024+0 Datensätze aus > 1073741824 Bytes (1,1 GB) kopiert, 14,0558 s, 76,4 MB/s > > > So this is read, this is what write looks like: > > > [root@sonde04 ~]# sync ; time { dd if=/dev/zero of=/tmp/test.dd bs=1M > count=1024 ; sync ; } > 1024+0 Datensätze ein > 1024+0 Datensätze aus > 1073741824 Bytes (1,1 GB) kopiert, 47,0048 s, 22,8 MB/s > > real0m50.428s > user0m0.022s > sys 0m7.021s > > > This is 1 GB (power of 2) written in 50 seconds, so that amounts to > about 20,5 MB/s. Would anyone be so kind and run the commands himself, > maybe with Debian or something with a more recent kernel. > > Also i would appreciate any tips on how to get a bit more write > performance. We would like to monitor traffic with it, using NtopNG and > tcpdump. When we dump traffic, we don't write payloads, just headers, so > we don't need more than maybe 40 MB/s guess. > > > Have a nice day, > > Karsten Kruse -- Thomas Fjellstrom tho...@fjellstrom.ca ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
Re: [Soekris] Poor I/O performance
Karsten Kruse schrieb: > Would anyone be so kind and run the commands himself, > maybe with Debian or something with a more recent kernel. # hdparm -i /dev/sda /dev/sda: Model=KingSpec KSM-mSATA.5-016SJ, FwRev=111007, SerialNo=MSA0512030900037 Config={ Fixed } RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=0 BuffType=unknown, BuffSize=unknown, MaxMultSect=16, MultSect=16 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=31277232 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120} PIO modes: pio0 pio3 pio4 DMA modes: mdma0 mdma1 mdma2 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 AdvancedPM=yes: disabled (255) WriteCache=enabled Drive conforms to: Unspecified: ATA/ATAPI-4,5,6,7 * signifies the current active mode # time dd if=/dev/sda of=/dev/null bs=1M count=1024 1024+0 Datensätze ein 1024+0 Datensätze aus 1073741824 Bytes (1,1 GB) kopiert, 7,54117 s, 142 MB/s real0m7.545s user0m0.013s sys 0m4.488s # sync ; time { dd if=/dev/zero of=/tmp/test.dd bs=1M count=1024 ; sync ; } 1024+0 Datensätze ein 1024+0 Datensätze aus 1073741824 Bytes (1,1 GB) kopiert, 26,4151 s, 40,6 MB/s real0m29.974s user0m0.021s sys 0m12.297s # hdparm -i /dev/sdc /dev/sdc: Model=SAMSUNG HN-M101MBB, FwRev=2AR10001, SerialNo=S2R8J9AB703589 Config={ Fixed } RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4 BuffType=unknown, BuffSize=8192kB, MaxMultSect=16, MultSect=off CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=1953525168 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120} PIO modes: pio0 pio1 pio2 pio3 pio4 DMA modes: mdma0 mdma1 mdma2 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 udma6 AdvancedPM=yes: disabled (255) WriteCache=enabled Drive conforms to: unknown: ATA/ATAPI-0,1,2,3,4,5,6,7 * signifies the current active mode # time dd if=/dev/sdc of=/dev/null bs=1M count=1024 1024+0 Datensätze ein 1024+0 Datensätze aus 1073741824 Bytes (1,1 GB) kopiert, 10,2349 s, 105 MB/s real0m10.239s user0m0.017s sys 0m4.772s # sync ; time { dd if=/dev/zero of=/mnt/sdc/test.dd bs=1M count=1024 ; sync ; } 1024+0 Datensätze ein 1024+0 Datensätze aus 1073741824 Bytes (1,1 GB) kopiert, 17,72 s, 60,6 MB/s real0m20.513s user0m0.011s sys 0m14.525s # lspci | grep RAID 0d:00.0 RAID bus controller: Silicon Image, Inc. Device 0242 (rev 01) # uname -a Linux daemon 3.4.34-gentoo-net6501 #2 SMP Sun Mar 9 00:27:36 UTC 2014 i686 Genuine Intel(R) CPU @ 1.60GHz GenuineIntel GNU/Linux -- Regards *WOL*fgang *S*chricker ___ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech