Performance for RAID 1+0

These results are slow. See our home page for expected RAID 1+0 results, which were performed with 4 Toshiba 2TB drives. (not enterprise)

perform some raw throughput on each drive and watch Activity Monitor to see what each disk is doing on reads and writes. (A SoftRAID verify disk, or start a certify will give you the result of what each disk can do)

Each standard Toshiba 2TB should perform around 170MB/s I don't know what the expected is on the Enterprise version, but I would expect slightly higher.

That is were I would start.

RAID 1+0 can sustain any disk failure, or a failure on each "side" of the array.

Best practice is to certify each disk before use (can take 2-5 days for a 3 pass certify, depending on the drive)

Thanks for your response and good advice information.

I did as you suggested.

I unmounted all 4 disks/volumes.

I then ran a SoftRAID Verify on each disk.

First LaCie 2big enclosure
2big's disk 1 gave a solid read rate of 160 MB/s
2big's disk 2 gave a solid read rate of 165 MB/s

Second LaCie 2big enclosure
2big's disk 1 gave a solid read rate range of 160 to 180 MB/s with occasional dips to 135 MB/s
2big's disk 2 gave a solid read rate range of 150 to 190 MB/s with occasional dips to 120 MB/s

Could it be that the 2nd 2big disks behaving with occasional dips to 120 to 135 MB/s explain what you've called as being "slow" for a RAID 1+0 ?

I did read your Home page and obviously I have questions about the RAID 1+0 test results as they might compare with my setup.

1) Your test employed a 2013 MacPro6,1 6core, 3.5 GHz and 16 GB RAM (1866 MHz) computer. This has Thunderbolt-2 ports.
2) Your test used 3TB Toshiba disks
3) Your enclosures were OWC ThunderBay 4 units. These are using Thunderbolt 2 interface technology.
4) The OWC enclosure contains all 4 disks.
5) Your test used AJA (not sure what version although I doubt this makes any difference)
6) There's no mention of what Mac OS X version you had installed on the MacPro6,1 (Suspect it was Yosemite)

1) My test employed a MacBook Pro8,3 (17-inch, late 2011), Intel i7 2.5 GHz 4cores and 16GB RAM (1333 MHz) computer
2) My MacBook Pro8,3 has a single Thunderbolt 1 port
3) My 4 disks were 2TB 7200 rpm Toshiba enterprise disk units.
4) I used two LaCie 2big enclosures; each had two of the Toshiba disks in them.
5) I employed 4 of the Toshiba disks in two LaCie 2big enclosures that were daisy chained from the MacBook Pro8,3 using Thunderbolt -1 interface/technology
6) I used AJA as was stated in the Home page
7) My test Mac X OS was El Capitan 10.11.3

My guess is your tests are better than mine because
1) You're using a much faster computer than me; MP6,1 vs. MBP8,3
2) You're using Thunderbolt 2 interface/technology vs. my Thunderbolt 1 interface/technology
3) You're using a single Thunderbolt -2 capable enclosure containing all 4 disks vs. my two enclosures with 2 disks each and daisy chained using Thunderbolt 1 interface/technology

There's nothing I can do to use different Thunderbolt ports for each of my LaCie 2big enclosures as my MBP8,3 has but a single Thunderbolt 1 port. So daisy chaining is my only choice.

Thus I suspect my setup is slower on account of having my disks spread between two enclosures which are daisy chained together, using Thunderbolt 1, having a slower performing MBP8,3 computer and that two disk in my set of 4 behaved a bit strange with occasional dips in write rate of some -70 MB/s compared to their peak rates of around 190 MB/s.

Does my analysis make sense to you ?

As some further information....

My AJA test were conducted after having filled the disks to about 69% full. This can obviously make a big difference as AJA would be writing to the slower sections of the disks.

I made a few raw read of fairly large files that would be in the first 5% of the disk. The read rates were around 200 MB/s.

Given this I would expect to see read rates at around 350 MB/sec for my RAID 1+0. This still is unfortunately lower than your Home page results of 534 MB/sec, but for good reasons I suspect as I mentioned in my previous posting. Your MP6,1 6core computer, Thunderbolt 2, using the OWC ThunderboltBay 4 and RAID 1+0 obviously makes a big difference in the ultimate performance.

Thank you.

RAID 1+0 is not so susceptible to CPU for performance The drives in neither test came close to Thunderbolt throughput.

But you answered the question in your detailed questions.

Once you copy files to a volume, it slows down significantly. Your results make perfect sense if it was more than 50% full!

If you freshly erase your volume, then ran the tests, you would get much much better performance.

We ran AJA as:
16GB file size
Largest frame rate
Disable file system cache checkbox DISABLED (real world uses the file system cache)