ZFS Performance Testing: P4 Clone using WD Raptors

In my previous post on ZFS performance testing I ran through various tests on a particular test system that I had running at the time. This time round, we are testing a Intel motherboard with a Pentium 4 CPU, 2GB of RAM, and 6x74GB drives. This system was literally thrown together with spare parts laying around the home office. This is the ultimate in home made, commodity parts, down and dirty NAS setups. Using seven (6) Western Digital Raptor drives attached to three different SATA II non-RAID controllers. For the uninitiated, Raptors are SATA drives that spin at 10K RPM. So here we go….

*Note*
My intention was to run this with seven (7) disks so that it would be a WD Raptor vs WD Blue test. Unfortunately, my seventh, and last WD Raptor died during configuration. With that in mind, it is interesting nonetheless to compare the 7 disk WD Blue results with the 6 WD Raptor results.

Test rig:

• Custom Build
• Intel Pentium 4 @ 3.0GHz Hyper Threading Enabled
• 2GB RAM
• Onboard SATA II
• 6x74GB SATA II 10000 RPM – Six (6) independent drives with no RAID. Model: Western Digital Raptor
• FreeNAS 0.7.2 Sabanda (revision 5543)-ZFS v13

GNU dd:
Tests performed from the CLI using good ole’ GNU dd. The following command was used to first write, and then read back:

dd if=/dev/zero of=foo bs=2M count=10000 ; dd if=foo of=/dev/null bs=2M

Results:
Results are listed as configuration, write, then read.

• ZFS stripe pool utilizing six (6) SATA disks
  • 150 MB/s
  • 239 MB/s

• ZFS stripe pool utilizing six (6) SATA disks with dataset compression set to “On”
  • 302 MB/s
  • 515 MB/s

  • ZFS raidz pool utilizing six (6) SATA disks
    • 99 MB/s
    • 165 MB/s

  • ZFS raidz pool utilizing six (6) SATA disks with dataset compression set to “On”
    • 299 MB/s
    • 516 MB/s

  • ZFS raidz2 pool utilizing six (6) SATA disks
    • 76 MB/s
    • 164 MB/s

  • ZFS raidz2 pool utilizing six (6) SATA disks with dataset compression set to “On”
    • 301 MB/s
    • 514 MB/s

    Notes, Thoughts & Mentionables:
    There are a few things worth mentioning about this system:
    This is a truly down and dirty quick and ugly build using used parts. As such, you get what you pay for and the performance data here proves that. Possibly more influential to the performance scores here could and likely is the storage controllers. The motherboard only has four ports so used/cheap/old SATA II PCI controller was used to gain the additional three ports.

    As always, the tests involving compression provide interesting insight into the limitations of various processors. While running the compression off tests, the CPU load was relatively low and the systems audible sound was unchanged. While running compression on tests, the CPU was of course showing a heavy load, but it also prompted the CPU cooler to spin at a higher (more audible) rate. Guess those old P4’s still cook 🙂

    –himuraken