Is this transfer-speed correct?

[SnorreSelmer]

I'm wondering if I did something wrong when I replicated a speed-test someone else did elsewhere on the forum (some thread about having network speed-problems).

From what I have read about ZFS, running a raidz shouldn't give any speed-boost (n drives in raidz should give the speed of one drive), but when I ran the command below I got a 650MiB/sec write-speed!

Code:

dd if=/dev/zero of=/mnt/zpool/testfile.test bs=1024M count=8

Since I have six Seagate Barracuda Green 5900RPM drives (SATA 6Gbit/sec) in SATA-2 ports I guess I'm pretty much maxing the capacity of each drive completely, or is this a case of false numbers due to stuff being cached to RAM without my knowledge?

EDIT: Bummer! I know why the numbers are so high... I have lzjb compression turned on for my entire zpool... Ofcourse I'll get good speeds when I'm writing something as easily compressed as a bunch of zeroes. I'll make a new dataset with compression explicitly turned off and try again!

 

[Scine]

I'm wondering if I did something wrong when I replicated a speed-test someone else did elsewhere on the forum (some thread about having network speed-problems).

From what I have read about ZFS, running a raidz shouldn't give any speed-boost (n drives in raidz should give the speed of one drive), but when I ran the command below I got a 650MiB/sec write-speed!

Code:

dd if=/dev/zero of=/mnt/zpool/testfile.test bs=1024M count=8

Since I have six Seagate Barracuda Green 5900RPM drives (SATA 6Gbit/sec) in SATA-2 ports I guess I'm pretty much maxing the capacity of each drive completely, or is this a case of false numbers due to stuff being cached to RAM without my knowledge?

EDIT: Bummer! I know why the numbers are so high... I have lzjb compression turned on for my entire zpool... Ofcourse I'll get good speeds when I'm writing something as easily compressed as a bunch of zeroes. I'll make a new dataset with compression explicitly turned off and try again!

\

This is what I got:

8+0 records in
8+0 records out
8589934592 bytes transferred in 119.252520 secs (72031472 bytes/sec)

Around 70mb per second, so with multiply that out, and I think it's on the high side, but quite possible for you to get that speed.

 

[Milhouse]

HP N36L Microserver (dual-core 1.3GHz Atom-class CPU), 4GB RAM, FreeNAS 8.0.2-RELEASE AMD64, 4x1TB Samsung HD103SJ:

Code:

WRITE     RAIDZ1          RAIDZ2
512B      212,143,286     148,067,931
4K        211,027,799     145,136,615

READ      RAIDZ1          RAIDZ2
512B      134,556,578     101,815,843
4K        191,722,812     108,278,443
512B+RA   417,207,180     320,201,198
4K+RA	  422,350,077     341,636,028

512B=No 4K Sectors Forced
4K=4K Sectors Forced
RA=read-ahead enabled

Tests:

Code:

WRITE: dd if=/dev/zero of=/mnt/share/test.dat bs=1M count=6000
READ:  dd of=/dev/null if=/mnt/share/test.dat bs=1M count=6000

 

[survive]

Hi SnorreSelmer,

Be sure you are writing out enough data to fill up all the caches. I use something like this:

dd if=/dev/zero of=testfile bs=8192k count=2000

which writes out a 16GB file.

-Will