FreeBSD 8, AMD64, ZFS Storage server

Hi!

We are getting 2 new storage servers and have chosen to go the FreeBSD 8/ZFS route. We have looked into a lot of enterprise storage (Sun/NetApp/EMC/Equallogic), but just feel that the bang for the buck is quite low. We know of course that the service agreements on enterprise storage is in a whole different ballpark than what we are planning for. Anyway...

Just wanted to see if anyone has any type of similar setup or had any problems with the included hardware. We have planned on AMD 6-core CPU running x64 version of FreeBSD. But can also switch to Intel 6-core.

Motherboard: Supermicro H8DI3+-F

CPU: AMD 6-core Istanbul 2,2GHz

RAM: 8x2GB RAM or 8x4GB DDR2 RAM REG ECC

Diskcontroller: Adaptec RAID 51645 (20 port) w BBU

Disk Cache HW: Adaptec MaxIQ SSD kit

OS boot: 2 x 150GB WD Velociraptor 10K 2.5" drives

ZFS ZIL: 1x Intel X25-E SSD, 32GB SATAII, SLC

Harddrives: 16 x WD RE4 Green 2TB

Extra NIC: Intel Gigabit ET Quad port server adapter

Chassis: SuperMicro 3U storage server chassis

Please feel free to drop a post regarding the hardware and running FreeBSD 8 on AMD64 and using ZFS.
I have been searching the forums and will continue my search but any info is great.

Best regards to you all and have a nice New Year!

/Dave
 
Ditch the RAID controller if you are using ZFS for everything. Just pick up multiple 8-port controllers (LSI has some nice ones, fully supported in FreeBSD).

You don't need a single ~$1500 RAID controller, when you can use multiple $300 controllers, and put them on separate PCI-X/PCIe buses. One of the main goals of ZFS is to get away from massive, expensive, overly-redundant RAID controllers.

If you are going to split the ZIL onto a separate device, then you *MUST* make it a mirrored vdev. If the ZIL device ever dies, the entire pool goes with it!! ZFSv13 (in FreeBSD 8) doesn't support the removal of ZIL devices.

And, if you search the forums, you'll find my thread on rsbackup, which is our storage server setup using FreeBSD and ZFS. :)

You also don't need 10K RPM drives for the OS. In fact, you can put the entire FreeBSD OS (/ and /usr) onto 2 GB CompactFlash. The data drives should be the fast ones, not the OS drive. Especially since most of it will be in RAM once you boot. :) In fact, using CompactFlash or SD for the OS would give you more data storage space.

2x CF for the OS.
2x SSD for the ZIL.
18x SATA drives for storage.
 
I've heard that about the ZIL a few times, but i'm curious if this is confirmed or not.

According to the ZFS best practices wiki, while it IS impossible to remove a ZIL device once it has been added, it also says that in the event of ZIL device failure, ZIL function drops back to the main pool.

This is confusing....
 
phoenix
LSI has some nice ones, fully supported in FreeBSD

Could you suggest someone that is supported in 8.0-Release? What about 8204ELP?
Would like to make raid-z zpool via cheap LSI controller & Intel expanders (AXX6DRV3GEXP & AXX4DRV3GEXP)

P.S. MB is Intel S3000AH & Case SC5300
 
Here's the results of my research through various manufacturer websites, various online shopping sites, and various FreeBSD mailing list archives and man pages. Taken from an e-mail to our hardware tech.

Code:
The following cards are supported by FreeBSD and Linux (lots of good reviews
from FreeBSD users), and are in the $100-300 range.  Unfortunately, newegg.ca,
ncix.com, and cdw.com all show them as backordered.  :(

LSI SAS 3081E-R    8-port PCIe,  2 mini-SAS connectors
LSI SAS 3080X-R    8-port PCI-X, 2 mini-SAS connectors

SuperMicro AOC-USAS-L8i   8-port PCIe, 2 mini-SAS connectors
SuperMicro AOC-USASLP-L8i 8-port PCIe, 2 mini-SAS connectors
(these two are the same, one's just a low-profile version)

The SuperMicro boards are actually UIO cards, and not true PCIe cards.  They
work in PCIe slots, but require custom backplates in order to work in PCIe
slots.  They can be used without the backplates, though.

There's also the option of going with 8-port 3Ware cards, as they are in the
$400-$600 range, instead of over $1000 for the 12- and 16-port versions.

3Ware 9550SXU-8LP  8-port PCI-X, 8 SATA connectors (not ideal, ML is preferred)
3Ware 9650SE-8ML   8-port PCIe,  2 mini-SAS connectors

If LSI does decide to drop the 3Ware line of cards (really, really hope not),
the next best-supported RAID controllers for Linux/FreeBSD would probably be
Areca. These appear to be about 1/2 the price of the 3Ware controllers.

Areca ARC-1300ix-16  16-port PCIe, 4 mini-SAS connectors
(this is a non-RAID controller, and 8 of the ports are external, so should be treated as an 8-port card)

Areca ARC-1120        8-port PCI-X, 8 SATA connectors (not ideal)

Areca ARC-1130ML     12-port PCI-X, 3 ML-SATA connectors (same as 3Ware 9550)
Areca ARC-1160ML     16-port PCI-X, 4 ML-SATA connectors (same as 3Ware 9550)

Areca ARC-1222        8-port PCIe,  2 mini-SAS connectors
Areca ARC-1231ML     12-port PCIe,  3 mini-SAS connectors
 
My configuration under testing:

Mainboard: http://www.supermicro.com/Aplus/motherboard/Opteron2000/MCP55/H8DME-2.cfm
Raid Controller: 3Ware (now LSI) 9690SA SAS Contr.
Chassis: http://www.supermicro.com/products/chassis/4U/846/SC846E2-R900.cfm
NIC: Intel PCI-X dualport serveradapter

Raid1 with 2x WD2502ABYS 250GB
Raid6 with 10x WD1002FBYS 1 TB

"Performance" Setting for the Controller is set to "Balanced", means the cache on the disks itself should
not be used so the BBU can make his job saving the data on the controller cache in case of power failure.

The 9690SA-8I is compatible with the SAS expanders on the backplane of the chassis.
The SAS-846EL2 backplane has two expanders which allow effective failover and
recovery.

little problems:

3dm (3dm2) Port under FreeBSD seems to be to old for this Controller/Firmware Version = displayed the raid configuration wrong
the commandline tool tw_cli works
i had to upgrade the controller firmware because of the SAS-846EL2 backplane.

big problems:

had to disable the onboard NICs because they just gone away sometimes ... have the same problems with other onboard NICs on nvidia chipsets too.
well i dont care about it and just i added a dualport intel NIC. this will do a great job.

ZFS:

Code:
pool: home
state: ONLINE
scrub: none requested
config:

	NAME        STATE     READ WRITE CKSUM
	home        ONLINE       0     0     0
	  da1       ONLINE       0     0     0


Code:
NAME   SIZE   USED  AVAIL    CAP  HEALTH  ALTROOT
home  7.25T   479G  6.78T     6%  ONLINE  -

dmesg:

Code:
Copyright (c) 1992-2009 The FreeBSD Project.
Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
	The Regents of the University of California. All rights reserved.
FreeBSD is a registered trademark of The FreeBSD Foundation.
FreeBSD 8.0-RELEASE-p1 #1: Wed Dec  9 12:17:00 CET 2009
    root@host.domain.net:/usr/obj/usr/src/sys/FS
WARNING: WITNESS option enabled, expect reduced performance.
WARNING: DIAGNOSTIC option enabled, expect reduced performance.
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Dual-Core AMD Opteron(tm) Processor 2214 (2211.35-MHz K8-class CPU)
  Origin = "AuthenticAMD"  Id = 0x40f12  Stepping = 2
  Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT>
  Features2=0x2001<SSE3,CX16>
  AMD Features=0xea500800<SYSCALL,NX,MMX+,FFXSR,RDTSCP,LM,3DNow!+,3DNow!>
  AMD Features2=0x1f<LAHF,CMP,SVM,ExtAPIC,CR8>
real memory  = 34359738368 (32768 MB)
avail memory = 33164169216 (31627 MB)
ACPI APIC Table: <S M C  OEMAPIC >
FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs
FreeBSD/SMP: 2 package(s) x 2 core(s)
 cpu0 (BSP): APIC ID:  0
 cpu1 (AP): APIC ID:  1
 cpu2 (AP): APIC ID:  2
 cpu3 (AP): APIC ID:  3
ioapic0 <Version 1.1> irqs 0-23 on motherboard
kbd1 at kbdmux0
acpi0: <S M C OEMXSDT> on motherboard
acpi0: [ITHREAD]
acpi0: Power Button (fixed)
acpi0: reservation of fec00000, 1000 (3) failed
acpi0: reservation of fee00000, 1000 (3) failed
acpi0: reservation of 0, a0000 (3) failed
acpi0: reservation of 100000, dff00000 (3) failed
Timecounter "ACPI-fast" frequency 3579545 Hz quality 1000
acpi_timer0: <24-bit timer at 3.579545MHz> port 0x2008-0x200b on acpi0
pcib0: <ACPI Host-PCI bridge> port 0xcf8-0xcff on acpi0
pci0: <ACPI PCI bus> on pcib0
pci0: <memory, RAM> at device 0.0 (no driver attached)
isab0: <PCI-ISA bridge> at device 1.0 on pci0
isa0: <ISA bus> on isab0
ichsmb0: <SMBus controller> port 0xbc00-0xbc3f,0x2d00-0x2d3f,0x2e00-0x2e3f irq 21 at device 1.1 on pci0
ichsmb0: [ITHREAD]
smbus0: <System Management Bus> on ichsmb0
smb0: <SMBus generic I/O> on smbus0
ohci0: <OHCI (generic) USB controller> mem 0xfe8bf000-0xfe8bffff irq 22 at device 2.0 on pci0
ohci0: [ITHREAD]
usbus0: <OHCI (generic) USB controller> on ohci0
ehci0: <EHCI (generic) USB 2.0 controller> mem 0xfe8bec00-0xfe8becff irq 23 at device 2.1 on pci0
ehci0: [ITHREAD]
usbus1: EHCI version 1.0
usbus1: <EHCI (generic) USB 2.0 controller> on ehci0
pcib1: <ACPI PCI-PCI bridge> at device 6.0 on pci0
pci1: <ACPI PCI bus> on pcib1
vgapci0: <VGA-compatible display> port 0xc000-0xc0ff mem 0xf0000000-0xf7ffffff,0xfe9f0000-0xfe9fffff irq 16 at device 5.0 on pci1
pcib2: <ACPI PCI-PCI bridge> at device 10.0 on pci0
pci2: <ACPI PCI bus> on pcib2
pcib3: <ACPI PCI-PCI bridge> at device 0.0 on pci2
pci3: <ACPI PCI bus> on pcib3
pcib4: <ACPI PCI-PCI bridge> at device 0.1 on pci2
pci4: <ACPI PCI bus> on pcib4
pcib5: <ACPI PCI-PCI bridge> at device 13.0 on pci0
pci5: <ACPI PCI bus> on pcib5
pcib6: <ACPI PCI-PCI bridge> at device 14.0 on pci0
pci6: <ACPI PCI bus> on pcib6
3ware device driver for 9000 series storage controllers, version: 3.70.05.001
twa0: <3ware 9000 series Storage Controller> port 0xd800-0xd8ff mem 0xfc000000-0xfdffffff,0xfeaff000-0xfeafffff irq 17 at device 0.0 on pci6
twa0: [ITHREAD]
twa0: INFO: (0x04: 0x0053): Battery capacity test is overdue: 
twa0: INFO: (0x15: 0x1300): Controller details:: Model 9690SA-8I, 128 ports, Firmware FH9X 4.10.00.007, BIOS BE9X 4.08.00.002
pcib7: <ACPI PCI-PCI bridge> at device 15.0 on pci0
pci7: <ACPI PCI bus> on pcib7
em0: <Intel(R) PRO/1000 Network Connection 6.9.14> port 0xec00-0xec1f mem 0xfebe0000-0xfebfffff,0xfebc0000-0xfebdffff irq 18 at device 0.0 on pci7
em0: Using MSI interrupt
em0: [FILTER]
em0: Ethernet address: 00:15:17:d2:df:60
em1: <Intel(R) PRO/1000 Network Connection 6.9.14> port 0xe880-0xe89f mem 0xfeb80000-0xfeb9ffff,0xfeb60000-0xfeb7ffff irq 17 at device 0.1 on pci7
em1: Using MSI interrupt
em1: [FILTER]
em1: Ethernet address: 00:15:17:d2:df:61
amdtemp0: <AMD K8 Thermal Sensors> on hostb3
amdtemp1: <AMD K8 Thermal Sensors> on hostb7
acpi_button0: <Power Button> on acpi0
uart0: <16550 or compatible> port 0x3f8-0x3ff irq 4 flags 0x10 on acpi0
uart0: [FILTER]
uart1: <16550 or compatible> port 0x2f8-0x2ff irq 3 on acpi0
uart1: [FILTER]
atrtc0: <AT realtime clock> port 0x70-0x71 irq 8 on acpi0
cpu0: <ACPI CPU> on acpi0
powernow0: <PowerNow! K8> on cpu0
cpu1: <ACPI CPU> on acpi0
powernow1: <PowerNow! K8> on cpu1
cpu2: <ACPI CPU> on acpi0
powernow2: <PowerNow! K8> on cpu2
cpu3: <ACPI CPU> on acpi0
powernow3: <PowerNow! K8> on cpu3
orm0: <ISA Option ROMs> at iomem 0xc0000-0xcafff,0xcb000-0xccfff on isa0
sc0: <System console> at flags 0x100 on isa0
sc0: VGA <16 virtual consoles, flags=0x300>
vga0: <Generic ISA VGA> at port 0x3c0-0x3df iomem 0xa0000-0xbffff on isa0
atkbdc0: <Keyboard controller (i8042)> at port 0x60,0x64 on isa0
atkbd0: <AT Keyboard> irq 1 on atkbdc0
kbd0 at atkbd0
atkbd0: [GIANT-LOCKED]
atkbd0: [ITHREAD]
Timecounters tick every 1.000 msec
usbus0: 12Mbps Full Speed USB v1.0
usbus1: 480Mbps High Speed USB v2.0
ugen0.1: <nVidia> at usbus0
uhub0: <nVidia OHCI root HUB, class 9/0, rev 1.00/1.00, addr 1> on usbus0
ugen1.1: <nVidia> at usbus1
uhub1: <nVidia EHCI root HUB, class 9/0, rev 2.00/1.00, addr 1> on usbus1
uhub0: 10 ports with 10 removable, self powered
...
da0 at twa0 bus 0 target 0 lun 0
da0: <AMCC 9690SA-8I  DISK 4.10> Fixed Direct Access SCSI-5 device 
da0: 100.000MB/s transfers
da0: 238408MB (488259584 512 byte sectors: 255H 63S/T 30392C)
SMP: AP CPU #1 Launched!
SMP: AP CPU #2 Launched!
SMP: AP CPU #3 Launched!
WARNING: WITNESS option enabled, expect reduced performance.
WARNING: DIAGNOSTIC option enabled, expect reduced performance.
da1 at twa0 bus 0 target 1 lun 0
da1: <AMCC 9690SA-8I  DISK 4.10> Fixed Direct Access SCSI-5 device 
da1: 100.000MB/s transfers
da1: 7629312MB (15624830976 512 byte sectors: 255H 63S/T 972600C)
Root mount waiting for: usbus1
uhub1: 10 ports with 10 removable, self powered
Root mount waiting for: usbus1
ugen1.2: <Peppercon AG> at usbus1
ums0: <Peppercon AG Multidevice, class 0/0, rev 2.00/0.01, addr 2> on usbus1
ums0: 3 buttons and [Z] coordinates ID=0
ukbd0: <Peppercon AG Multidevice, class 0/0, rev 2.00/0.01, addr 2> on usbus1
kbd2 at ukbd0
Trying to mount root from ufs:/dev/da0s1a
ZFS filesystem version 13
ZFS storage pool version 13
 
iozone results for 31x 1GB simultanous

iozone 31x 1GB simultanous:

Code:
Record Size 8 KB
        File size set to 1048576 KB
Output is in Kbytes/sec
        Time Resolution = 0.000002 seconds.
        Processor cache size set to 1024 Kbytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
        Min process = 1 
        Max process = 64 
        Throughput test with 1 process
        Each process writes a 1048576 Kbyte file in 8 Kbyte records


Children see throughput for 31 initial writers  =  201265.78 KB/sec
        Parent sees throughput for 31 initial writers   =  150463.73 KB/sec
        Min throughput per process                      =    5986.81 KB/sec 
        Max throughput per process                      =    7588.98 KB/sec
        Avg throughput per process                      =    6492.44 KB/sec
        Min xfer                                        =  827280.00 KB

        Children see throughput for 31 rewriters        =  118843.87 KB/sec
        Parent sees throughput for 31 rewriters         =  115923.02 KB/sec
        Min throughput per process                      =    3547.24 KB/sec 
        Max throughput per process                      =    4006.70 KB/sec
        Avg throughput per process                      =    3833.67 KB/sec
        Min xfer                                        =  928512.00 KB

        Children see throughput for 31 readers          =  345760.32 KB/sec
        Parent sees throughput for 31 readers           =  340252.11 KB/sec
        Min throughput per process                      =    9941.60 KB/sec 
        Max throughput per process                      =   11947.37 KB/sec
        Avg throughput per process                      =   11153.56 KB/sec
        Min xfer                                        =  876416.00 KB

        Children see throughput for 31 re-readers       =  332146.74 KB/sec
        Parent sees throughput for 31 re-readers        =  327672.41 KB/sec
        Min throughput per process                      =    4804.59 KB/sec 
        Max throughput per process                      =   20229.93 KB/sec
        Avg throughput per process                      =   10714.41 KB/sec
        Min xfer                                        =  252032.00 KB

        Children see throughput for 31 reverse readers  =  348791.73 KB/sec
        Parent sees throughput for 31 reverse readers   =  340139.45 KB/sec
        Min throughput per process                      =    1201.33 KB/sec 
        Max throughput per process                      =   14732.41 KB/sec
        Avg throughput per process                      =   11251.35 KB/sec
        Min xfer                                        =   87552.00 KB

        Children see throughput for 31 stride readers   =   23292.33 KB/sec
        Parent sees throughput for 31 stride readers    =   23276.54 KB/sec
        Min throughput per process                      =     595.24 KB/sec 
        Max throughput per process                      =     855.33 KB/sec
        Avg throughput per process                      =     751.37 KB/sec
        Min xfer                                        =  730592.00 KB

        Children see throughput for 31 random readers   =    8350.64 KB/sec
        Parent sees throughput for 31 random readers    =    8350.49 KB/sec
        Min throughput per process                      =     267.10 KB/sec 
        Max throughput per process                      =     271.54 KB/sec
        Avg throughput per process                      =     269.38 KB/sec
        Min xfer                                        = 1031432.00 KB

        Children see throughput for 31 mixed workload   =    5781.48 KB/sec
        Parent sees throughput for 31 mixed workload    =    5369.59 KB/sec
        Min throughput per process                      =     174.16 KB/sec 
        Max throughput per process                      =     198.01 KB/sec
        Avg throughput per process                      =     186.50 KB/sec
        Min xfer                                        =  922288.00 KB

        Children see throughput for 31 random writers   =    3412.19 KB/sec
        Parent sees throughput for 31 random writers    =    3372.15 KB/sec
        Min throughput per process                      =     108.75 KB/sec 
        Max throughput per process                      =     111.33 KB/sec
        Avg throughput per process                      =     110.07 KB/sec
        Min xfer                                        = 1024200.00 KB

        Children see throughput for 31 pwrite writers   =  169670.93 KB/sec
        Parent sees throughput for 31 pwrite writers    =  134318.70 KB/sec
        Min throughput per process                      =    4719.29 KB/sec 
        Max throughput per process                      =    6281.94 KB/sec
        Avg throughput per process                      =    5473.26 KB/sec
        Min xfer                                        =  787928.00 KB

        Children see throughput for 31 pread readers    =  360126.34 KB/sec
        Parent sees throughput for 31 pread readers     =  353519.89 KB/sec
        Min throughput per process                      =   10226.21 KB/sec 
        Max throughput per process                      =   12556.96 KB/sec
        Avg throughput per process                      =   11616.98 KB/sec
        Min xfer                                        =  876416.00 KB
 
User23 said:
...
Raid1 with 2x WD2502ABYS 250GB
Raid6 with 10x WD1002FBYS 1 TB
...
Code:
	NAME        STATE     READ WRITE CKSUM
	home        ONLINE       0     0     0
	  da1       ONLINE       0     0     0

...
Code:
NAME   SIZE   USED  AVAIL    CAP  HEALTH  ALTROOT
home  7.25T   479G  6.78T     6%  ONLINE  -
...

Why are you using the RAID-controllers RAID-functionality if you are dumping ZFS on top? Any special reasoning? Why not plug the disks straight into ZFS.
 
Well ZFS raidz1/2 does a good job, but it needs much more cpu power.
I would need 8 or 12 cores to get the same performance.

I had to use this controller because of the backplane and it is the only one i have used for a long time without problems. So i know what i have to expect. And if you have (must use) a real hardware raid controller why you want to spend CPU time for a software raid on it?

Why ZFS on top of the raid6?
To use all the other features ZFS brings with it and be more flexible in the future.
And ofc UFS2 cant handle such a big array :)

I hope the NFS performance of ZFS is not as low as some reports say. lets see ...

---

For example:

I have the following confguration in use for one year now. first it was only one raidz1 with 6x 1TB. later i added a second controller and 6x 2 TB in a second raidz1 to the tank.

Under heavy disk io the cpu is the bottleneck. The performance is great anyway ... no question :)

Intel Q6600 (4x 2,4GHz, 2x 4MB cache)
8GB RAM
2x 3ware 9550 8x SATA (as stable ATA controller)(the 2 free sata ports on each controller could be used later for ZIL devices)

Code:
zpool status
  pool: backup1
 state: ONLINE
 scrub: none requested
config:

	NAME        STATE     READ WRITE CKSUM
	backup1     ONLINE       0     0     0
	  raidz1    ONLINE       0     0     0
	    da0     ONLINE       0     0     0
	    da1     ONLINE       0     0     0
	    da2     ONLINE       0     0     0
	    da3     ONLINE       0     0     0
	    da4     ONLINE       0     0     0
	    da5     ONLINE       0     0     0
	  raidz1    ONLINE       0     0     0
	    da6     ONLINE       0     0     0
	    da7     ONLINE       0     0     0
	    da8     ONLINE       0     0     0
	    da9     ONLINE       0     0     0
	    da10    ONLINE       0     0     0
	    da11    ONLINE       0     0     0
 
User23 said:
"Performance" Setting for the Controller is set to "Balanced", means the cache on the disks itself should
not be used so the BBU can make his job saving the data on the controller cache in case of power failure.

If you have a BBU installed, then set this to Performance. Since you have a BBU, you can enable the drive caches, and use the controller cache as well.

If you don't have a BBU, but you do have a good, working UPS configured to do an ordered shutdown, you can also set this to Performance and enable the drive caches.

If you don't have a BBU, but plan on using ZFS, then you can also set this to Performance and enable the drive caches. Also, if using ZFS, don't use the hardware RAID features. Just create "Single Drive" arrays for each disk attached to the controller. Then use the individual drives to create the zpool.

3dm (3dm2) Port under FreeBSD seems to be to old for this Controller/Firmware Version = displayed the raid configuration wrong

Uninstall the port version, and install the FreeBSD package off the CD. It installs into /usr and /etc instead of /usr/local, but it works. There's a whole slew of extra, fancy new features (like Power Consumption and drive Temperature data).

i had to upgrade the controller firmware because of the SAS-846EL2 backplane.

That's a good idea, regardless of the backplane or drives in use. :)

had to disable the onboard NICs because they just gone away sometimes ... have the same problems with other onboard NICs on nvidia chipsets too.
well i dont care about it and just i added a dualport intel NIC. this will do a great job.

nvidia NIC and harddrive chipsets are hit-and-miss in FreeBSD. Disabling all the onboard stuff that you don't use is best.


See above. Using ZFS with a single hardware RAID array like this eliminates 90% of the usefulness of ZFS. It can't do any redundancy checking or self-healing.
 
User23 said:
iozone results for 31x 1GB simultanous

Do the same tests, after removing the hardware raid array, and using the individual drives to create the zpool, using multiple raidz (or raidz2) vdevs.
 
User23 said:
Well ZFS raidz1/2 does a good job, but it needs much more cpu power.
I would need 8 or 12 cores to get the same performance.

Have you actually done any testing to prove that? ;)

I had to use this controller because of the backplane and it is the only one i have used for a long time without problems. So i know what i have to expect. And if you have (must use) a real hardware raid controller why you want to spend CPU time for a software raid on it?

When you get so much more than just "software RAID", heck yes. :)

Without access to multiple disks, ZFS is pretty much useless. It can't do any self-healing, it can't do any behind-the-scenes data corruption checking/fixing, it can't fix any corrupted files, it can't stripe data across multiple drives for improved performance.

In fact, the only thing you get by using ZFS on a single device is quick-and-easy FS creation, and snapshots. That's less than 10% of the features of ZFS.

This can quickly lead to a corrupted pool, with unrecoverable files.
 
phoenix said:
If you have a BBU installed, then set this to Performance. Since you have a BBU, you can enable the drive caches, and use the controller cache as well.

The BBU cant save the on disk cache. In worst case you will loose all the data thats currently waits in disk cache for writing. Except ZFS will not use the disk cache for writing ... i did not know that.

phoenix said:
If you don't have a BBU, but you do have a good, working UPS configured to do an ordered shutdown, you can also set this to Performance and enable the drive caches.

It is more a religion decission :) . I saw failing UPS in the past with my own eyes.
 
phoenix said:
Have you actually done any testing to prove that? ;)

It is in progress ;P

Code:
  pool: home
 state: ONLINE
 scrub: none requested
config:

	NAME        STATE     READ WRITE CKSUM
	home        ONLINE       0     0     0
	  raidz2    ONLINE       0     0     0
	    da1     ONLINE       0     0     0
	    da2     ONLINE       0     0     0
	    da3     ONLINE       0     0     0
	    da4     ONLINE       0     0     0
	    da5     ONLINE       0     0     0
	    da6     ONLINE       0     0     0
	    da7     ONLINE       0     0     0
	    da8     ONLINE       0     0     0
	    da9     ONLINE       0     0     0
	    da10    ONLINE       0     0     0

First results: Load 9.x+ while writing

Code:
        Children see throughput for 31 initial writers  =  161468.34 KB/sec
        Parent sees throughput for 31 initial writers   =  114066.57 KB/sec
        Min throughput per process                      =    4521.95 KB/sec 
        Max throughput per process                      =    6521.20 KB/sec
        Avg throughput per process                      =    5208.66 KB/sec
        Min xfer                                        =  727216.00 KB

Ill posts the full report after its done

When you get so much more than just "software RAID", heck yes. :)

Yes it consumes CPU time like hell...

Without access to multiple disks, ZFS is pretty much useless. It can't do any self-healing, it can't do any behind-the-scenes data corruption checking/fixing, it can't fix any corrupted files, it can't stripe data across multiple drives for improved performance.

In fact, the only thing you get by using ZFS on a single device is quick-and-easy FS creation, and snapshots. That's less than 10% of the features of ZFS.

It is the same problem like: http://hub.opensolaris.org/bin/view/Community+Group+zfs/faq#HDoesZFSworkwithSANattacheddevices

This can quickly lead to a corrupted pool, with unrecoverable files.

This sounds like ZFS is self destructing.
 
User23 said:
It is in progress ;P

Code:
  pool: home
 state: ONLINE
 scrub: none requested
config:

	NAME        STATE     READ WRITE CKSUM
	home        ONLINE       0     0     0
	  raidz2    ONLINE       0     0     0
	    da1     ONLINE       0     0     0
	    da2     ONLINE       0     0     0
	    da3     ONLINE       0     0     0
	    da4     ONLINE       0     0     0
	    da5     ONLINE       0     0     0
	    da6     ONLINE       0     0     0
	    da7     ONLINE       0     0     0
	    da8     ONLINE       0     0     0
	    da9     ONLINE       0     0     0
	    da10    ONLINE       0     0     0
Wouldn't 2 raidz1 be faster and recommended by sun?
"The recommended number of disks per group is between 3 and 9. If you have more disks, use multiple groups."
 
Matty said:
User23 said:
It is in progress ;P
Wouldn't 2 raidz1 be faster and recommended by sun?
"The recommended number of disks per group is between 3 and 9. If you have more disks, use multiple groups."

True, ill test that 2x raidz1 with 5 disk config and a raidz2 config with 9 disks ... like it is recommended.
 
So, ZFS (in a single 9-drive raidz2) takes a 20% performance hit ... but gives you so much more in return.

Will be interesting to see how the multiple raidz vdevs setups perform in comparison. With 9 drives, you can use the following setups:
  • 1x 9-drive raidz2 vdev
  • 2x 4-drive raidz2 vdevs
  • 3x 3-drive raidz1 vdevs
 
Hi again!
We've had a lot of holidays here in Sweden so I haven't followed this thread until today... :-D

Well we went in a bit of a different direction:
Motherboard: Supermicro H8DI3+-F
CPU: AMD 6-core Istanbul 2,2GHz
RAM: 8x4GB DDR2 RAM REG ECC
Diskcontroller: 2 x Adaptec RAID 51245 (12 port) w BBU (BBU is supposed to give better perf on the card)(We will not use RAID from the cards)
OS boot: 2 x 150GB WD Velociraptor (internal disks)
ZFS ZIL: 2 x Intel X25-E SSD, 32GB SATAII (one read, one write)
Harddrives: 22 x WD RE4 Green 2TB
Extra NIC: Intel Gigabit ET Quad port server adapter
Chassis: SuperMicro 4U storage server chassis
OS: FreeBSD 8

We're planning on getting two of these so we can replicate or something between volumes, between the storage systems. We just not sure how this is done yet with BSD/ZFS compared to for example NetApp with its replication license.

And once again thanks for the input! I'll be coming with lots of more questions in the future.

/Dave
 
You can use ZFS send/recv to transfer snapshots between the two servers. Depending on how much data is being written/changed, you can auto-create snapshots every minutes, 10 minutes, hour, etc, and send them to the other server. Quite a few Solaris sites do this to "replicate" the data on two servers.

Rsync works as well (that's what we went with, since ZFS recv in ZFSv6 on FreeBSD couldn't do more than a couple dozen KBps).
 
phoenix said:
You can use ZFS send/recv to transfer snapshots between the two servers. Depending on how much data is being written/changed, you can auto-create snapshots every minutes, 10 minutes, hour, etc, and send them to the other server. Quite a few Solaris sites do this to "replicate" the data on two servers.

Rsync works as well (that's what we went with, since ZFS recv in ZFSv6 on FreeBSD couldn't do more than a couple dozen KBps).

Thanks! Will look into this. We have mostly talked about using rsync but we're not sure this would work either.

We are going to firstly use the storage for VMWare servers from NFS on ZFS (this is what we have planned on). We have 2 identical servers and want to be able to fail over between the two if needed.
I'm just a bit unsure how we will be able to "snapshot" the VMWare servers VMDK files in a easy way while the servers are live.
I know with NetApp you can either create a pretty complex script routine or buy the software that does all the bits for you and some. We hope we will be able to do something similar here, otherwise we'll have to go with more traditional methods.

/Dave
 
phoenix said:
So, ZFS (in a single 9-drive raidz2) takes a 20% performance hit ... but gives you so much more in return.

Will be interesting to see how the multiple raidz vdevs setups perform in comparison. With 9 drives, you can use the following setups:
  • 1x 9-drive raidz2 vdev
  • 2x 4-drive raidz2 vdevs
  • 3x 3-drive raidz1 vdevs

Yes iam interested in too. It will take some days to test that.

It looks like ZFS could perform better than the raidcontroller if configured like recommended :) . Except for writing, the load is always under 2.

Code:
						    ZFS on single hw raid6 device	ZFS 1x raidz2		ZFS 1x raidz2
							10 disks (not recom.)	      10 disks (not recom.)	  9 disks
								7,25TB			     7,08TB		  6,2TB

        Children see throughput for 31 initial writers  =  201265.78 KB/sec		161468.34 KB/sec	194090.46 KB/sec
        Parent sees throughput for 31 initial writers   =  150463.73 KB/sec		114066.57 KB/sec	 96803.54 KB/sec		
        Min throughput per process                      =    5986.81 KB/sec 		  4521.95 KB/sec	  5342.48 KB/sec	
        Max throughput per process                      =    7588.98 KB/sec		  6521.20 KB/sec	 11659.86 KB/sec	
        Avg throughput per process                      =    6492.44 KB/sec		  5208.66 KB/sec	  6260.98 KB/sec		
        Min xfer                                        =  827280.00 KB			727216.00 KB		519184.00 KB
 
Code:
						    ZFS on single hw raid6 device	ZFS 1x raidz2		ZFS 1x raidz2		ZFS 2x raidz2
							10 disks (not recom.)	      10 disks (not recom.)	  9 disks		  2x5 disks 
								7,25TB			     7,08TB		  6,2TB			    5,28TB 

        Children see throughput for 31 initial writers  =  201265.78 KB/sec		161468.34 KB/sec	194090.46 KB/sec	118971.07 KB/sec
        Parent sees throughput for 31 initial writers   =  150463.73 KB/sec		114066.57 KB/sec	 96803.54 KB/sec	 71035.95 KB/sec
        Min throughput per process                      =    5986.81 KB/sec 		  4521.95 KB/sec	  5342.48 KB/sec	  3111.14 KB/sec
        Max throughput per process                      =    7588.98 KB/sec		  6521.20 KB/sec	 11659.86 KB/sec	  6005.53 KB/sec
        Avg throughput per process                      =    6492.44 KB/sec		  5208.66 KB/sec	  6260.98 KB/sec	  3837.78 KB/sec
        Min xfer                                        =  827280.00 KB			727216.00 KB		519184.00 KB		543256.00 KB

        Children see throughput for 31 rewriters        =  118843.87 KB/sec		 88296.12 KB/sec	 94851.35 KB/sec	 87543.58 KB/sec
        Parent sees throughput for 31 rewriters         =  115923.02 KB/sec		 86382.02 KB/sec	 92564.80 KB/sec	 85627.08 KB/sec
        Min throughput per process                      =    3547.24 KB/sec 		  2628.55 KB/sec	  2859.38 KB/sec	  2586.37 KB/sec
        Max throughput per process                      =    4006.70 KB/sec		  3024.69 KB/sec	  3225.33 KB/sec	  3072.57 KB/sec
        Avg throughput per process                      =    3833.67 KB/sec		  2848.26 KB/sec	  3059.72 KB/sec	  2823.99 KB/sec
        Min xfer                                        =  928512.00 KB			911360.00 KB		929664.00 KB		882688.00 KB

        Children see throughput for 31 readers          =  345760.32 KB/sec		355439.88 KB/sec	310864.48 KB/sec	280303.74 KB/sec
        Parent sees throughput for 31 readers           =  340252.11 KB/sec		347992.79 KB/sec	302392.68 KB/sec	274431.44 KB/sec
        Min throughput per process                      =    9941.60 KB/sec 		  2929.78 KB/sec	  5941.07 KB/sec	  4764.28 KB/sec
        Max throughput per process                      =   11947.37 KB/sec              13671.29 KB/sec         12443.30 KB/sec	 10151.20 KB/sec
        Avg throughput per process                      =   11153.56 KB/sec		 11465.80 KB/sec	 10027.89 KB/sec	  9042.06 KB/sec
        Min xfer                                        =  876416.00 KB			228480.00 KB		515968.00 KB		507776.00 KB

        Children see throughput for 31 re-readers       =  332146.74 KB/sec		365836.72 KB/sec	319717.53 KB/sec	285893.18 KB/sec
        Parent sees throughput for 31 re-readers        =  327672.41 KB/sec		356374.37 KB/sec	309249.20 KB/sec	278385.91 KB/sec
        Min throughput per process                      =    4804.59 KB/sec 		  4327.40 KB/sec	  4874.84 KB/sec	  5189.78 KB/sec
        Max throughput per process                      =   20229.93 KB/sec		 14428.39 KB/sec	 18179.50 KB/sec	 11341.52 KB/sec
        Avg throughput per process                      =   10714.41 KB/sec		 11801.18 KB/sec	 10313.47 KB/sec	  9222.36 KB/sec
        Min xfer                                        =  252032.00 KB			316288.00 KB		286592.00 KB		483200.00 KB

        Children see throughput for 31 reverse readers  =  348791.73 KB/sec		378207.58 KB/sec	297222.73 KB/sec	294168.65 KB/sec
        Parent sees throughput for 31 reverse readers   =  340139.45 KB/sec		367847.12 KB/sec	285330.94 KB/sec	283241.22 KB/sec
        Min throughput per process                      =    1201.33 KB/sec 		  8882.71 KB/sec	    50.86 KB/sec	  7071.22 KB/sec
        Max throughput per process                      =   14732.41 KB/sec		 15323.43 KB/sec	 18723.45 KB/sec	 11208.89 KB/sec
        Avg throughput per process                      =   11251.35 KB/sec		 12200.24 KB/sec	  9587.83 KB/sec	  9489.31 KB/sec
        Min xfer                                        =   87552.00 KB			614272.00 KB		  2944.00 KB		679808.00 KB

        Children see throughput for 31 stride readers   =   23292.33 KB/sec		 19494.66 KB/sec	 19303.88 KB/sec	 17640.89 KB/sec
        Parent sees throughput for 31 stride readers    =   23276.54 KB/sec		 19478.48 KB/sec	 19293.39 KB/sec	 17627.54 KB/sec
        Min throughput per process                      =     595.24 KB/sec 		   440.73 KB/sec	   544.24 KB/sec	   544.73 KB/sec
        Max throughput per process                      =     855.33 KB/sec		   940.35 KB/sec	   729.74 KB/sec	   578.91 KB/sec
        Avg throughput per process                      =     751.37 KB/sec		   628.86 KB/sec	   622.71 KB/sec	   569.06 KB/sec
        Min xfer                                        =  730592.00 KB			491480.00 KB		782112.00 KB		986928.00 KB

        Children see throughput for 31 random readers   =    8350.64 KB/sec		  3139.91 KB/sec	  3185.64 KB/sec	  6372.45 KB/sec
        Parent sees throughput for 31 random readers    =    8350.49 KB/sec		  3139.90 KB/sec	  3185.62 KB/sec	  6372.39 KB/sec
        Min throughput per process                      =     267.10 KB/sec 		    99.68 KB/sec	   100.98 KB/sec	   202.37 KB/sec
        Max throughput per process                      =     271.54 KB/sec		   105.75 KB/sec	   104.69 KB/sec	   208.62 KB/sec
        Avg throughput per process                      =     269.38 KB/sec		   101.29 KB/sec	   102.76 KB/sec	   205.56 KB/sec
        Min xfer                                        = 1031432.00 KB			988416.00 KB	       1011392.00 KB	       1017176.00 KB

        Children see throughput for 31 mixed workload   =    5781.48 KB/sec		  2543.61 KB/sec	  2476.89 KB/sec	  4219.44 KB/sec
        Parent sees throughput for 31 mixed workload    =    5369.59 KB/sec		  2444.68 KB/sec	  2401.83 KB/sec	  4052.52 KB/sec
        Min throughput per process                      =     174.16 KB/sec		    78.96 KB/sec	    77.61 KB/sec           131.44 KB/sec
        Max throughput per process                      =     198.01 KB/sec		    85.42 KB/sec	    82.78 KB/sec	   141.04 KB/sec
        Avg throughput per process                      =     186.50 KB/sec		    82.05 KB/sec	    79.90 KB/sec	   136.11 KB/sec
        Min xfer                                        =  922288.00 KB			969232.00 KB		983040.00 KB		977208.00 KB

        Children see throughput for 31 random writers   =    3412.19 KB/sec		  1957.64 KB/sec	  1941.00 KB/sec	  2526.20 KB/sec
        Parent sees throughput for 31 random writers    =    3372.15 KB/sec		  1937.07 KB/sec	  1928.09 KB/sec	  2500.80 KB/sec
        Min throughput per process                      =     108.75 KB/sec 		    62.68 KB/sec	    62.19 KB/sec	    80.77 KB/sec
        Max throughput per process                      =     111.33 KB/sec		    63.80 KB/sec	    63.01 KB/sec	    82.33 KB/sec
        Avg throughput per process                      =     110.07 KB/sec		    63.15 KB/sec	    62.61 KB/sec	    81.49 KB/sec
        Min xfer                                        = 1024200.00 KB		       1030208.00 KB	       1034928.00 KB	       1028768.00 KB

        Children see throughput for 31 pwrite writers   =  169670.93 KB/sec		127665.01 KB/sec	187936.90 KB/sec	107282.57 KB/sec
        Parent sees throughput for 31 pwrite writers    =  134318.70 KB/sec		 89630.82 KB/sec	152674.50 KB/sec	 63974.87 KB/sec
        Min throughput per process                      =    4719.29 KB/sec 		  3410.34 KB/sec	  5262.72 KB/sec	  2586.47 KB/sec
        Max throughput per process                      =    6281.94 KB/sec		  5516.22 KB/sec	  6980.33 KB/sec	  5281.52 KB/sec
        Avg throughput per process                      =    5473.26 KB/sec		  4118.23 KB/sec          6062.48 KB/sec	  3460.73 KB/sec
        Min xfer                                        =  787928.00 KB			646424.00 KB		795136.00 KB		511616.00 KB

        Children see throughput for 31 pread readers    =  360126.34 KB/sec		338863.82 KB/sec	297671.57 KB/sec	282012.07 KB/sec
        Parent sees throughput for 31 pread readers     =  353519.89 KB/sec		330358.98 KB/sec	289853.82 KB/sec        273999.37 KB/sec
        Min throughput per process                      =   10226.21 KB/sec 		  8784.75 KB/sec          6710.06 KB/sec	  6983.21 KB/sec
        Max throughput per process                      =   12556.96 KB/sec		 12120.25 KB/sec	 12898.82 KB/sec	 11158.35 KB/sec
        Avg throughput per process                      =   11616.98 KB/sec		 10931.09 KB/sec          9602.31 KB/sec	  9097.16 KB/sec
        Min xfer                                        =  876416.00 KB			786432.00 KB		548992.00 KB		679808.00 KB

raidz1 and mirror results will follow
 
raidz1 with 9 drives 7.11TB

Code:
Excel chart generation enabled
	Record Size 8 KB
	File size set to 1048576 KB
	Command line used: iozone -R -l 31 -u 31 -r 8k -s 1024m -F /home/f0 /home/f1 /home/f2 /home/f3 /home/f4 /home/f5 /home/f6 /home/f7 
/home/f8 /home/f9 /home/f10 /home/f11 /home/f12 /home/f13 /home/f14 /home/f15 /home/f16 /home/f17 /home/f18 /home/f19 /home/f20 /home/f21 
/home/f22 /home/f23 /home/f24 /home/f25 /home/f26 /home/f27 /home/f28 /home/f29 /home/f30 /home/f31 /home/f32 /home/f33 /home/f34 /home /f35 
/home/f36 /home/f37 /home/f38 /home/f39 /home/f40 /home/f41 /home/f42 /home/f43 /home/f44 /home/f45 /home/f46 /home/f47 /home/f48 /home/f49 
/home/f50 /home/f51 /home/f52 /home/f53 /home/f54 /home/f55 /home/f56 /home/f57 /home/f58 /home/f59 /home/f60 /home/f61 /home/f62 /home/f63

	Output is in Kbytes/sec
	Time Resolution = 0.000002 seconds.
	Processor cache size set to 1024 Kbytes.
	Processor cache line size set to 32 bytes.
	File stride size set to 17 * record size.
	Min process = 31 
	Max process = 31 
	Throughput test with 31 processes
	Each process writes a 1048576 Kbyte file in 8 Kbyte records

	Children see throughput for 31 initial writers 	=  208172.21 KB/sec
	Parent sees throughput for 31 initial writers 	=  172595.51 KB/sec
	Min throughput per process 			=    5782.47 KB/sec 
	Max throughput per process 			=    7616.98 KB/sec
	Avg throughput per process 			=    6715.23 KB/sec
	Min xfer 					=  796976.00 KB

	Children see throughput for 31 rewriters 	=  102606.86 KB/sec
	Parent sees throughput for 31 rewriters 	=  101171.05 KB/sec
	Min throughput per process 			=    3014.62 KB/sec 
	Max throughput per process 			=    3511.05 KB/sec
	Avg throughput per process 			=    3309.90 KB/sec
	Min xfer 					=  900352.00 KB

	Children see throughput for 31 readers 		=  387198.20 KB/sec
	Parent sees throughput for 31 readers 		=  379311.31 KB/sec
	Min throughput per process 			=    5154.57 KB/sec 
	Max throughput per process 			=   15272.38 KB/sec
	Avg throughput per process 			=   12490.26 KB/sec
	Min xfer 					=  355456.00 KB

	Children see throughput for 31 re-readers 	=  320033.11 KB/sec
	Parent sees throughput for 31 re-readers 	=  310994.17 KB/sec
	Min throughput per process 			=    5401.99 KB/sec 
	Max throughput per process 			=   25386.70 KB/sec
	Avg throughput per process 			=   10323.65 KB/sec
	Min xfer 					=  226304.00 KB

	Children see throughput for 31 reverse readers 	=  305630.74 KB/sec
	Parent sees throughput for 31 reverse readers 	=  293534.26 KB/sec
	Min throughput per process 			=    1712.60 KB/sec 
	Max throughput per process 			=   20117.15 KB/sec
	Avg throughput per process 			=    9859.06 KB/sec
	Min xfer 					=   90752.00 KB

	Children see throughput for 31 stride readers 	=   22283.38 KB/sec
	Parent sees throughput for 31 stride readers 	=   22273.16 KB/sec
	Min throughput per process 			=     479.02 KB/sec 
	Max throughput per process 			=     871.18 KB/sec
	Avg throughput per process 			=     718.82 KB/sec
	Min xfer 					=  576664.00 KB

	Children see throughput for 31 random readers 	=    2874.69 KB/sec
	Parent sees throughput for 31 random readers 	=    2874.68 KB/sec
	Min throughput per process 			=      90.81 KB/sec 
	Max throughput per process 			=      95.65 KB/sec
	Avg throughput per process 			=      92.73 KB/sec
	Min xfer 					=  995608.00 KB

	Children see throughput for 31 mixed workload 	=    2278.92 KB/sec
	Parent sees throughput for 31 mixed workload 	=    2271.76 KB/sec
	Min throughput per process 			=      72.73 KB/sec 
	Max throughput per process 			=      74.09 KB/sec
	Avg throughput per process 			=      73.51 KB/sec
	Min xfer 					= 1029408.00 KB

	Children see throughput for 31 random writers 	=    1965.92 KB/sec
	Parent sees throughput for 31 random writers 	=    1951.16 KB/sec
	Min throughput per process 			=      63.14 KB/sec 
	Max throughput per process 			=      63.88 KB/sec
	Avg throughput per process 			=      63.42 KB/sec
	Min xfer 					= 1036416.00 KB

	Children see throughput for 31 pwrite writers 	=  159855.55 KB/sec
	Parent sees throughput for 31 pwrite writers 	=  127850.95 KB/sec
	Min throughput per process 			=    4240.99 KB/sec 
	Max throughput per process 			=    6337.74 KB/sec
	Avg throughput per process 			=    5156.63 KB/sec
	Min xfer 					=  701960.00 KB

	Children see throughput for 31 pread readers 	=  368381.51 KB/sec
	Parent sees throughput for 31 pread readers 	=  357053.00 KB/sec
	Min throughput per process 			=    8721.74 KB/sec 
	Max throughput per process 			=   14701.04 KB/sec
	Avg throughput per process 			=   11883.27 KB/sec
	Min xfer 					=  638976.00 KB


2x raidz1 with 2x5 disks 7.13TB

Code:
Record Size 8 KB
	File size set to 1048576 KB
	Command line used: iozone -R -l 31 -u 31 -r 8k -s 1024m -F /home/f0 /home/f1 /home/f2 /home/f3 /home/f4 /home/f5 /home/f6 /home/f7 
/home/f8 /home/f9 /home/f10 /home/f11 /home/f12 /home/f13 /home/f14 /home/f15 /home/f16 /home/f17 /home/f18 /home/f19 /home/f20 /home/f21 
/home/f22 /home/f23 /home/f24 /home/f25 /home/f26 /home/f27 /home/f28 /home/f29 /home/f30 /home/f31 /home/f32 /home/f33 /home/f34 /home /f35 
/home/f36 /home/f37 /home/f38 /home/f39 /home/f40 /home/f41 /home/f42 /home/f43 /home/f44 /home/f45 /home/f46 /home/f47 /home/f48 /home/f49 
/home/f50 /home/f51 /home/f52 /home/f53 /home/f54 /home/f55 /home/f56 /home/f57 /home/f58 /home/f59 /home/f60 /home/f61 /home/f62 /home/f63

	Output is in Kbytes/sec
	Time Resolution = 0.000002 seconds.
	Processor cache size set to 1024 Kbytes.
	Processor cache line size set to 32 bytes.
	File stride size set to 17 * record size.
	Min process = 31 
	Max process = 31 
	Throughput test with 31 processes
	Each process writes a 1048576 Kbyte file in 8 Kbyte records

	Children see throughput for 31 initial writers 	=  240188.71 KB/sec
	Parent sees throughput for 31 initial writers 	=  181504.13 KB/sec
	Min throughput per process 			=    7062.25 KB/sec 
	Max throughput per process 			=    8601.47 KB/sec
	Avg throughput per process 			=    7748.02 KB/sec
	Min xfer 					=  861184.00 KB

	Children see throughput for 31 rewriters 	=  113480.85 KB/sec
	Parent sees throughput for 31 rewriters 	=  111265.15 KB/sec
	Min throughput per process 			=    3483.71 KB/sec 
	Max throughput per process 			=    3888.38 KB/sec
	Avg throughput per process 			=    3660.67 KB/sec
	Min xfer 					=  939520.00 KB

	Children see throughput for 31 readers 		=  307583.62 KB/sec
	Parent sees throughput for 31 readers 		=  299550.20 KB/sec
	Min throughput per process 			=    4702.65 KB/sec 
	Max throughput per process 			=   18115.81 KB/sec
	Avg throughput per process 			=    9922.05 KB/sec
	Min xfer 					=  278912.00 KB

	Children see throughput for 31 re-readers 	=  374780.61 KB/sec
	Parent sees throughput for 31 re-readers 	=  364328.27 KB/sec
	Min throughput per process 			=    4937.61 KB/sec 
	Max throughput per process 			=   16304.53 KB/sec
	Avg throughput per process 			=   12089.70 KB/sec
	Min xfer 					=  325504.00 KB

	Children see throughput for 31 reverse readers 	=  384149.35 KB/sec
	Parent sees throughput for 31 reverse readers 	=  374955.97 KB/sec
	Min throughput per process 			=    5885.06 KB/sec 
	Max throughput per process 			=   15980.62 KB/sec
	Avg throughput per process 			=   12391.91 KB/sec
	Min xfer 					=  390912.00 KB

	Children see throughput for 31 stride readers 	=   23573.38 KB/sec
	Parent sees throughput for 31 stride readers 	=   23556.19 KB/sec
	Min throughput per process 			=     692.92 KB/sec 
	Max throughput per process 			=     803.08 KB/sec
	Avg throughput per process 			=     760.43 KB/sec
	Min xfer 					=  904880.00 KB

	Children see throughput for 31 random readers 	=    5180.75 KB/sec
	Parent sees throughput for 31 random readers 	=    5180.72 KB/sec
	Min throughput per process 			=     164.68 KB/sec 
	Max throughput per process 			=     168.72 KB/sec
	Avg throughput per process 			=     167.12 KB/sec
	Min xfer 					= 1023472.00 KB

	Children see throughput for 31 mixed workload 	=    4229.80 KB/sec
	Parent sees throughput for 31 mixed workload 	=    4049.69 KB/sec
	Min throughput per process 			=     131.69 KB/sec 
	Max throughput per process 			=     142.48 KB/sec
	Avg throughput per process 			=     136.45 KB/sec
	Min xfer 					=  969168.00 KB

	Children see throughput for 31 random writers 	=    3364.70 KB/sec
	Parent sees throughput for 31 random writers 	=    3322.79 KB/sec
	Min throughput per process 			=     107.54 KB/sec 
	Max throughput per process 			=     109.82 KB/sec
	Avg throughput per process 			=     108.54 KB/sec
	Min xfer 					= 1026848.00 KB

	Children see throughput for 31 pwrite writers 	=  179130.74 KB/sec
	Parent sees throughput for 31 pwrite writers 	=   80142.70 KB/sec
	Min throughput per process 			=    4533.41 KB/sec 
	Max throughput per process 			=   11170.48 KB/sec
	Avg throughput per process 			=    5778.41 KB/sec
	Min xfer 					=  425600.00 KB

	Children see throughput for 31 pread readers 	=  365544.69 KB/sec
	Parent sees throughput for 31 pread readers 	=  359612.10 KB/sec
	Min throughput per process 			=      42.71 KB/sec 
	Max throughput per process 			=   13506.08 KB/sec
	Avg throughput per process 			=   11791.76 KB/sec
	Min xfer 					=    3328.00 KB
 
3x raidz1 3x3 disks 5.35T

Code:
Record Size 8 KB
	File size set to 1048576 KB
	Command line used: iozone -R -l 31 -u 31 -r 8k -s 1024m -F /home/f0 /home/f1 /home/f2 /home/f3 /home/f4 /home/f5 /home/f6 /home/f7 
/home/f8 /home/f9 /home/f10 /home/f11 /home/f12 /home/f13 /home/f14 /home/f15 /home/f16 /home/f17 /home/f18 /home/f19 /home/f20 /home/f21 
/home/f22 /home/f23 /home/f24 /home/f25 /home/f26 /home/f27 /home/f28 /home/f29 /home/f30 /home/f31 /home/f32 /home/f33 /home/f34 /home /f35 
/home/f36 /home/f37 /home/f38 /home/f39 /home/f40 /home/f41 /home/f42 /home/f43 /home/f44 /home/f45 /home/f46 /home/f47 /home/f48 /home/f49 
/home/f50 /home/f51 /home/f52 /home/f53 /home/f54 /home/f55 /home/f56 /home/f57 /home/f58 /home/f59 /home/f60 /home/f61 /home/f62 /home/f63

	Output is in Kbytes/sec
	Time Resolution = 0.000002 seconds.
	Processor cache size set to 1024 Kbytes.
	Processor cache line size set to 32 bytes.
	File stride size set to 17 * record size.
	Min process = 31 
	Max process = 31 
	Throughput test with 31 processes
	Each process writes a 1048576 Kbyte file in 8 Kbyte records

	Children see throughput for 31 initial writers 	=  194977.18 KB/sec
	Parent sees throughput for 31 initial writers 	=  145238.19 KB/sec
	Min throughput per process 			=    5577.41 KB/sec 
	Max throughput per process 			=    7678.55 KB/sec
	Avg throughput per process 			=    6289.59 KB/sec
	Min xfer 					=  761736.00 KB

	Children see throughput for 31 rewriters 	=  107264.31 KB/sec
	Parent sees throughput for 31 rewriters 	=  105691.77 KB/sec
	Min throughput per process 			=    3106.30 KB/sec 
	Max throughput per process 			=    3759.90 KB/sec
	Avg throughput per process 			=    3460.14 KB/sec
	Min xfer 					=  866304.00 KB

	Children see throughput for 31 readers 		=  329907.42 KB/sec
	Parent sees throughput for 31 readers 		=  325495.85 KB/sec
	Min throughput per process 			=    9172.33 KB/sec 
	Max throughput per process 			=   11446.83 KB/sec
	Avg throughput per process 			=   10642.17 KB/sec
	Min xfer 					=  843648.00 KB

	Children see throughput for 31 re-readers 	=  333976.38 KB/sec
	Parent sees throughput for 31 re-readers 	=  323647.98 KB/sec
	Min throughput per process 			=    7145.07 KB/sec 
	Max throughput per process 			=   14048.41 KB/sec
	Avg throughput per process 			=   10773.43 KB/sec
	Min xfer 					=  548736.00 KB

	Children see throughput for 31 reverse readers 	=  348859.71 KB/sec
	Parent sees throughput for 31 reverse readers 	=  339545.68 KB/sec
	Min throughput per process 			=     153.09 KB/sec 
	Max throughput per process 			=   14841.71 KB/sec
	Avg throughput per process 			=   11253.54 KB/sec
	Min xfer 					=   10880.00 KB

	Children see throughput for 31 stride readers 	=   21995.84 KB/sec
	Parent sees throughput for 31 stride readers 	=   21976.20 KB/sec
	Min throughput per process 			=     618.73 KB/sec 
	Max throughput per process 			=     818.59 KB/sec
	Avg throughput per process 			=     709.54 KB/sec
	Min xfer 					=  792664.00 KB

	Children see throughput for 31 random readers 	=    7795.43 KB/sec
	Parent sees throughput for 31 random readers 	=    7795.35 KB/sec
	Min throughput per process 			=     249.06 KB/sec 
	Max throughput per process 			=     254.20 KB/sec
	Avg throughput per process 			=     251.47 KB/sec
	Min xfer 					= 1027400.00 KB

	Children see throughput for 31 mixed workload 	=    4849.54 KB/sec
	Parent sees throughput for 31 mixed workload 	=    4673.46 KB/sec
	Min throughput per process 			=     151.49 KB/sec 
	Max throughput per process 			=     161.06 KB/sec
	Avg throughput per process 			=     156.44 KB/sec
	Min xfer 					=  986280.00 KB

	Children see throughput for 31 random writers 	=    2961.60 KB/sec
	Parent sees throughput for 31 random writers 	=    2934.99 KB/sec
	Min throughput per process 			=      94.59 KB/sec 
	Max throughput per process 			=      96.31 KB/sec
	Avg throughput per process 			=      95.54 KB/sec
	Min xfer 					= 1029824.00 KB

	Children see throughput for 31 pwrite writers 	=  161924.70 KB/sec
	Parent sees throughput for 31 pwrite writers 	=  116085.19 KB/sec
	Min throughput per process 			=    4450.95 KB/sec 
	Max throughput per process 			=    6703.67 KB/sec
	Avg throughput per process 			=    5223.38 KB/sec
	Min xfer 					=  693496.00 KB

	Children see throughput for 31 pread readers 	=  306150.73 KB/sec
	Parent sees throughput for 31 pread readers 	=  300004.69 KB/sec
	Min throughput per process 			=    7810.73 KB/sec 
	Max throughput per process 			=   12104.51 KB/sec
	Avg throughput per process 			=    9875.83 KB/sec
	Min xfer 					=  679936.00 KB

5x mirror 5x2 disks 4.46T

Code:
Record Size 8 KB
	File size set to 1048576 KB
	Command line used: iozone -R -l 31 -u 31 -r 8k -s 1024m -F /home/f0 /home/f1 /home/f2 /home/f3 /home/f4 /home/f5 /home/f6 /home/f7 
/home/f8 /home/f9 /home/f10 /home/f11 /home/f12 /home/f13 /home/f14 /home/f15 /home/f16 /home/f17 /home/f18 /home/f19 /home/f20 /home/f21 
/home/f22 /home/f23 /home/f24 /home/f25 /home/f26 /home/f27 /home/f28 /home/f29 /home/f30 /home/f31 /home/f32 /home/f33 /home/f34 /home /f35 
/home/f36 /home/f37 /home/f38 /home/f39 /home/f40 /home/f41 /home/f42 /home/f43 /home/f44 /home/f45 /home/f46 /home/f47 /home/f48 /home/f49 
/home/f50 /home/f51 /home/f52 /home/f53 /home/f54 /home/f55 /home/f56 /home/f57 /home/f58 /home/f59 /home/f60 /home/f61 /home/f62 /home/f63

	Output is in Kbytes/sec
	Time Resolution = 0.000002 seconds.
	Processor cache size set to 1024 Kbytes.
	Processor cache line size set to 32 bytes.
	File stride size set to 17 * record size.
	Min process = 31 
	Max process = 31 
	Throughput test with 31 processes
	Each process writes a 1048576 Kbyte file in 8 Kbyte records

	Children see throughput for 31 initial writers 	=  165517.31 KB/sec
	Parent sees throughput for 31 initial writers 	=  113314.35 KB/sec
	Min throughput per process 			=    4822.53 KB/sec 
	Max throughput per process 			=    7166.15 KB/sec
	Avg throughput per process 			=    5339.27 KB/sec
	Min xfer 					=  705792.00 KB

	Children see throughput for 31 rewriters 	=  110261.11 KB/sec
	Parent sees throughput for 31 rewriters 	=  107713.00 KB/sec
	Min throughput per process 			=    3149.57 KB/sec 
	Max throughput per process 			=    4134.87 KB/sec
	Avg throughput per process 			=    3556.81 KB/sec
	Min xfer 					=  798720.00 KB

	Children see throughput for 31 readers 		=  298561.78 KB/sec
	Parent sees throughput for 31 readers 		=  289834.54 KB/sec
	Min throughput per process 			=    9011.21 KB/sec 
	Max throughput per process 			=   10670.23 KB/sec
	Avg throughput per process 			=    9631.03 KB/sec
	Min xfer 					=  885760.00 KB

	Children see throughput for 31 re-readers 	=  300249.48 KB/sec
	Parent sees throughput for 31 re-readers 	=  293567.68 KB/sec
	Min throughput per process 			=    8516.27 KB/sec 
	Max throughput per process 			=   11470.81 KB/sec
	Avg throughput per process 			=    9685.47 KB/sec
	Min xfer 					=  781568.00 KB

	Children see throughput for 31 reverse readers 	=  319173.17 KB/sec
	Parent sees throughput for 31 reverse readers 	=  309450.49 KB/sec
	Min throughput per process 			=     153.64 KB/sec 
	Max throughput per process 			=   13751.18 KB/sec
	Avg throughput per process 			=   10295.91 KB/sec
	Min xfer 					=   11776.00 KB

	Children see throughput for 31 stride readers 	=   20836.42 KB/sec
	Parent sees throughput for 31 stride readers 	=   20820.73 KB/sec
	Min throughput per process 			=     650.30 KB/sec 
	Max throughput per process 			=     690.90 KB/sec
	Avg throughput per process 			=     672.14 KB/sec
	Min xfer 					=  987072.00 KB

	Children see throughput for 31 random readers 	=   11431.42 KB/sec
	Parent sees throughput for 31 random readers 	=   11431.29 KB/sec
	Min throughput per process 			=     365.59 KB/sec 
	Max throughput per process 			=     372.43 KB/sec
	Avg throughput per process 			=     368.76 KB/sec
	Min xfer 					= 1029344.00 KB

	Children see throughput for 31 mixed workload 	=    8038.54 KB/sec
	Parent sees throughput for 31 mixed workload 	=    7542.01 KB/sec
	Min throughput per process 			=     245.69 KB/sec 
	Max throughput per process 			=     273.78 KB/sec
	Avg throughput per process 			=     259.31 KB/sec
	Min xfer 					=  940976.00 KB

	Children see throughput for 31 random writers 	=    5404.38 KB/sec
	Parent sees throughput for 31 random writers 	=    5326.68 KB/sec
	Min throughput per process 			=     172.30 KB/sec 
	Max throughput per process 			=     176.74 KB/sec
	Avg throughput per process 			=     174.33 KB/sec
	Min xfer 					= 1022264.00 KB

	Children see throughput for 31 pwrite writers 	=  110918.72 KB/sec
	Parent sees throughput for 31 pwrite writers 	=   41819.30 KB/sec
	Min throughput per process 			=    2468.24 KB/sec 
	Max throughput per process 			=    9939.37 KB/sec
	Avg throughput per process 			=    3578.02 KB/sec
	Min xfer 					=  259600.00 KB

	Children see throughput for 31 pread readers 	=  300026.61 KB/sec
	Parent sees throughput for 31 pread readers 	=  291201.70 KB/sec
	Min throughput per process 			=    8681.60 KB/sec 
	Max throughput per process 			=   11120.08 KB/sec
	Avg throughput per process 			=    9678.28 KB/sec
	Min xfer 					=  843648.00 KB
 
dejvnull said:

There is no such disk as WD RE4 Green, its either WD RE4 (which you linked to) or WD Green: http://westerndigital.com/en/products/Products.asp?DriveID=773

EDIT:

@dejvnull

I must appologize, I did not knew that RE-GP version exists, my bad :/
 
Back
Top