Raid 5 terrible performance with Thunderbay 4

Hi, I recently purchased a thunderbay 4 and getting terrible performance. My write speeds in Raid 5, or Raid 4 is only 100-120 MB no matter what I try.

The setup is consisting in 4 brand new seagate barracuda 7200 3 TB drives.

I have tried it in three different computers getting exactly the same results. 500+ read speeds, but only 100-120 write speeds. This is slower than the speed of a single drive when setting up as JBOD

If setup as RAID 0, I´m getting the expected 700 MB Read and close to 700 MB Write speed.

Raid 10 is giving me the expected 350 Mb write speed

All this measured with AJA, and BlackMagic, doesn't matter.

MODERATOR REPLIES:

To clarify this thread:
If the write cache for a RAID 4/5/10 volume is disabled, write performance will be much slower. There are two ways for write cache to be disabled:
A) Expired Trial mode disables the write cache.
B) Disabling Write Cache in the SoftRAID RAID preferences
(After a 30 day trial expires, the write cache is disabled)

Write Cache is important for write performance in RAID 4/5, as parity must be calculated in real time, and without caching this data into reserved RAM (until it can be written out to disk), performance will be poor on RAID 4/5 volumes.

A reason to intentionally disable Write Cache would be to make your volume more reliable if you are subject to power outages (get a UPS), brownouts (get a power conditioner, always a recommended option to make your computers/gear last longer), or if your computer is crashing for inexplicable reasons.
When your computer crashes, the volume's directory may be slightly damaged, as data that OS X thinks was sent to the disk was not actually written to the disks. Disabling the Write Cache on RAID volumes minimizes this problem, as all data is written directly to disk, rather than cached until the disks are less busy. However, this has an impact on write performance.

Other than a disabled Write Cache, there is no reason for RAID performance numbers not to match our published figures.

If your performance differs significantly, first ensure your volume has no data. Otherwise you are performing an invalid comparison, especially if your volume is full, or if data on the volumes is fragmented (scattered around the volume).

Other reasons are perhaps there is a hardware problem, or your drives/subsystems are significantly slower than what we used in our testing. (we generally use standard, off-the-shelf Thunderbay 4 enclosures from OWC, with standard Toshiba or Hitachi disk drives.) We do not use "customized" configurations to improve our numbers, any one should be able to achieve these with their standard hardware.

USB will always be much slower than Thunderbolt enclosures, USB is not really suited for high performance RAID application.

Note also that a Mac Pro will provide slightly better numbers than slower computers, or computers with less cores.

Hope this information clarifies this subject.

Feel free to keep discussing!