Last Updated on July 1, 2023 4:51 pm
July 1, 2023
I am posting this here but you can also view video showing the results (YouTube Video of SMR Test Results), or on my Reddit post here: REDDIT POST TO SMR RESULTS. There is basically the same content whether in this blog, the reddit post, or video, except I go into more detail explaining what SMR is in the video. There is a lot of data here and did my best to keep it concise and organized. There’s so many ways to slice it, but feel free to digest it however you see fit. I don’t blame you if you don’t want to hear my nasally voice ramble on for an hour.
This started as a simple curiosity experiment but ballooned into a much larger test set and took much longer than expected. Through multiple iterations, troubleshooting, learning new things, tests run as time permitted, took me well over a year to complete.
While most people here probably know what SMR is, I’ll just give my two cent summary.
Shingled Magnetic Recording technology, abbreviated as SMR, allows for disk drives to store more capacity per platter than a traditional hard drive. The technology is intended to reduce costs, because more data per platter means fewer platters and read/write heads. But because of how they store data the tradeoff is that write performance may degrade over time after enough data has been written to, deleted and then overwritten on the disk.
SMR drives are really intended for archival data, or data that is not frequently deleted or changed. However, since Western Digital decided to change their bottom tier WD RED NAS line of disk from CMR to SMR a few years back without full disclosure, many users were not happy with this change, as it had adverse effects greatly increasing rebuild times primarily of ZFS arrays. This got me wondering how much of an effect it really has in a parity RAID situation in other configurations.
DRIVES and TEST CONFIGS:
Three different SMR disks were tested to see how they would fare in popular RAID configurations.
The SMR test disks were all 2TB 3.5″ SATA drives:
- Seagate Barracuda Compute ST2000DM008
- WD Red WD20EFAX
- WD Blue WD20EZAZ
Several control CMR test disks were also utilized. The control set of disks were Seagate ST2000DM001 2TB 7200 RPM hard drives. The following 2TB CMR disks were also tested for comparison:
- Seagate Barracuda ST2000DM001
- WD Red Plus WD20EFZX
- Seagate Skyhawk ST2000VX008
Why were 2TB disks used? Bottom line, time and money. Some disks were already on hand, others were purchased solely for this test. Ultimately a Seagate Exos drive would have been preferable instead of Skyhawk, but at time of purchase, the Skyhawk was appreciably less expensive ($30 vs $80). 2TB also is the smallest capacity in 3.5″ form factor that comes in SMR as well as CMR. I also didn’t want to have to fill up more data than needed because writing multiple TB’s of data through dozens of tests is time consuming enough. Anything more than 2TB would have extended the test time considerably.
The following configurations were tested for single disk REBUILD times:
- OMV mdadm 4 disk RAID 5
- OMV ZFS 4 disk RAID Z1
- OMV SnapRAID Data & Parity disk
- Synology 4 disk SHR-1
- QNAP TR-004 4 disk “hardware” RAID 5
- UnRAID Data & Parity Disk
- Linux EXT4 Single Disk with and without TRIM
- Windows NTFS Single Disk with and without TRIM
- Controller: LSI 9211-8i vs Marvell 9215 vs Intel H77 MDADM RAID 5
- Controller: LSI 9211-8i vs Marvell 9215 vs Intel H77 ZFS RAID Z1
The configurations that were tested both WRITE and READ:
- NTFS Windows 10 Single Disk – tested over local SATA
- EXT4 OMV Single Disk – tested over 1GbE SMB
- XFS UnRAID single parity – tested over 1GbE SMB
- MDADM 4x 2TB RAID 5 OMV, formatted EXT4, tested over 1GbE SMB
- ZFS 4x 2TB RAID Z1 OMV, formatted ZFS (of course), tested over 1GbE SMB
- Synology BTRFS/MDADM 4x 2TB RAID 5 with DS920+, formatted BTRFS, tested over 1GbE SMB
- QNAP TR-004 DAS “Hardware” 4x 2TB RAID 5, formatted NTFS, tested over USB
HARDWARE SETUP
I made use of older hardware I had on hand. The PC config that was used for software NAS setup (i.e. OMV, UnRAID):
- CPU: Intel Core i5-3570 3.4GHz 4 core / 4 thread
- OS SSD: Sandisk Extreme 240GB
- Motherboard: Asus P8H77-I
- RAM: 2x8GB DDR3 1600
- PSU for PC: Seasonic S12II 430W 80plus Bronze
- PSU for HDD’s: FSP Group FSP270-60LE 270W
- SATA Controller: LSI 9211-8i
- LAN: Onboard Realtek RTL8111F Gigabit
- Disk Rack: Sans Digital HDD Rack 5 (https://www.sansdigital.com/hddrack5.html)
Note: the onboard SATA controller (Intel H77 Express) and also PCIe SATA controller (Marvell 9215) were utilized to compare performance between controllers in a few scenarios
The PC that ran the test programs to perform send and receive of file over ethernet and USB was configured as:
- OS: Windows 10 Pro
- Test Files SSD: Muskin Reactor 1TB 2.5″ SATA (READ FROM SSD WRITE TO TEST ARRAY)
- Test Files SSD: Crucial MX500 1TB 2.5″ SATA (WRITE TO SSD FROM TEST ARRAY)
- CPU: Intel Xeon E5-2630 v3 2.4GHz 8 core / 16 thread
- Motherboard: ASRock X99 Extreme4/3.1
- RAM: 2x16GB DDR4 1866 ECC
- SATA controller: onboard Intel X99
- LAN: onboard Intel 1218V 1GbE
- USB: onboard USB 3.1 5Gbps
TEST METHODOLOGY
Disks were not formatted or wiped between tests. All data from previous tests were left on the disk. Prior to each test, the disk was “cleared” or “formatted” using the minimal recommend process. For example in OMV, just running a “quick wipe” before adding the disk to a RAID 5 array.
All default settings for setting up an array were used. No TRIM or DISCARD settings were manually implemented except for the few specific TRIM tests.
For RAID array tests, 4x ST2000DM001 hard drives were used as control level to start, build array, add filler data, and then complete WRITE/READ test. Then one ST2000DM001 disk was then replaced with a test disk, rebuild the array, and then initiate the next read/write test for that disk. Subsequent tests would just swap out the test disk with another test disk, rebuild, and perform write/read testing until all disks were tested for that particular array. So 3x ST2000DM001 disks would always remain in the array with the fourth disk being a test disk.
For each initial array setup, the array was filled with random size and content data leaving approximately 1200 to 1500 GB free on a RAID array (out of ~ 6GB in a four disk single parity setup), approximately 60-70% filled capacity. This is considered the filler data and is not touched after the initial write to the disk or array.
For single disk tests, the filler data was filled to about 60-70% capacity, leaving about 800GB free for testing.
Sample random file distribution:
Total Files: 9979
File Size Min: 3KB
File Size Max: 2098998KB (~ 2.1GB)
File Size Avg: 154379 (~ 154MB)
% Files < 1MB: 1.74%
% Files 1MB to 10MB: 9.59%
% Files 10MB to 100MB: 45.59%
% Files 100MB to 1GB: 41.48%
% Files > 1GB: 1.60%
All test files were written from an SSD, and an actual file copy was executed from the Windows PC SSD over 1GbE or USB (depending on the test) using a Powershell script.
For RAID tests, gigabit ethernet was used because it is a typical use case to copy from a PC over SMB to a NAS. Any performance issues due to SMR should result in performance well below 1GbE speeds of about 112 MB/sec (realistic speeds).
A Powershell script was written and utilized to automate the tasks of file fill and measuring the per file transfer performance of the test data.
Random data was generated using RNGCryptoServiceProvider:
$rnd10 = (Get-Random 10) + 1
$rndmax = [int64]((Get-Random (2GB - 1))/$rnd10)
$bytes = (Get-Random $rndmax)
[System.Security.Cryptography.RNGCryptoServiceProvider] $rng = New-Object System.Security.Cryptography.RNGCryptoServiceProvider
$rndbytes = New-Object byte[] $bytes
$rng.GetBytes($rndbytes)
File transfer performance was measured using the StopWatch
command:
$StopWatch=[system.diagnostics.stopwatch]::startnew()
Copy-Item "$spath\$f\$file" "$dpath\TEST\$f"
$SecondsElapsed=$StopWatch.Elapsed.TotalSeconds
$StopWatch.Stop()
The average transfer speeds were computed using total size of files copied divided by summation of all time elapsed from above script.
Setting up a proper test to mitigate skewed results due to regular fragmentation was of high consideration. The following two test scenarios were devised which should highlight any effects of SMR degradation while minimizing any effect of fragmentation:
Test 1: Mixed Size Write / Read
- Write random size/content data to remaining full capacity of disk/array from Windows test PC.
- Use “delete” command to delete all random data just written.
- Write a test set of 620GB: 20x 10GB, 200x 1GB, 2000x 100MB, 20000x 1MB files from SSD and measure results per file.
- Read back half of files and measure results.
- Use “delete” command to delete test set of 620GB from test array.
- Immediately start next Scenario.
Test 2: Alternating 10MB / 1GB files
- Write 10MB size files to remaining full capacity of disk/array from Windows test PC.
- Use “delete” command to delete EVERY OTHER 10MB file, so half the 10MB files just written.
- Write 800x 1GB sized files to disk/array from SSD and measure results per file.
- Read back half of files and measure results per file.
- Use “delete” command to delete all 10MB and 1GB test files.
- Remove disk and start next disk test.
- Shut down, remove test disk and replace with next test disk.
Both scenarios take the disk to its fullest capacity and then immediately delete and then write data back to the disk with no idle time.
The second scenario should really bring forward any issues pertaining to SMR. Alternating 10MB file deletion should remove big chunks of data from each SMR zone that would require a rewrite of data to fill each SMR zone back.
BASELINE PERFORMANCE
Every disk was subjected to several baseline performance tests to ensure they were performing as intended:
- CrystalDiskMark 5x 1GB test set
- ATTO 512b to 64MB I/O with 1GB file size
- Hard Disk Sentinel Full Disk WRITE and READ
The CrystalDiskMark and ATTO benchmarks are omitted from this post, only because compiling the results will be quite tedious and honestly not very value added. However, I may provide them as raw image data in a future update to this blog.
That being said, below are results from the Hard Disk Sentinel Full Disk WRITE and READ tests for reference, which are probably more relevant anyhow.
The two SSD’s used to send and receive data were tested with HD Sentinel with results shown below:
- Test SSD read from to the test array Mushkin Reactor 1TB 2.5″ SATA
- Test SSD write to from the test array Crucial MX500 1TB 2.5″ SATA
FAILURES
Throughout testing, three of five WD Red WD20EFAX (SMR) had failures, all at different times in the testing. One became completely non responsive, it would power up but was never detected. Another started having increased failing sectors. The third one would fail out of a ZFS RAID Z1 rebuild despite having clean SMART and no other apparent issues. It was exchanged in hopes that the replacement would fare better. Each disk was successfully RMA’d promptly with a replacement within two weeks.
An SG Barracuda Compute ST2000DM008 (SMR) had to be RMA’d for being non-responsive and would hang the system periodically.
Each problematic disk went through a thorough troubleshooting process, changing PC’s, changing power supplies, changing cables, validating PSU voltages, etc before submitting for RMA. In each case, the replacement disk solved the issue completely (with the exception of the one instance with ZFS rebuild errors). Is this a result of SMR? Bad luck? Maybe some unknown other issue? I don’t know, but each of the replacement disks soldiered on through the rest of the testing without a hitch.
The SG Barracuda Compute ST2000DM008 (SMR) was not recognized as a device supporting TRIM despite in Linux desite it running TRIM just fine with NTFS in Windows. This is why there was an N/A TRIM in the TRIM results for the Seagate Barracuda Compute ST2000DM008 drive.
There were four WD RED Disks utilized. Three were WD20EFAX-68B2RN1 and one was WD20EFAX-68FB5N0. The 68FB5N0 would not TRIM in Linux despite it running TRIM just fine in NTFS Windows. 68B2RN1 drives would accept TRIM commands fine in Linux and Windows. This is why the TRIM test only included 3x WD RED instead of 4x WD RED.
So here are the results:
LEGEND:
SR = SnapRAID, SYNO = Synology DS920+, HW = Hardware
#### REBUILD TIMES:
**** RAID REBUILD TIMES (TIME IN MINUTES):
+XX% = % More time taken to complete than ST2000DM001 (worse)
-XX% = % Less time taken to complete than ST2000DM001 (better)
SR EXT4 BEFORE FILL = adding new test disk to pool and recreating 600GB of fill data from parity
SR EXT4 AFTER FILL = run fill disk test, delete all data, then recreate 600GB fill data from parity
CMR CMR CMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX
------------- ------------- ------------- ------------- ----------- -------------
MDADM RAID 5 205 226 (+10%) 242 (-18%) 205 ( 0%) 253 (+23%) 268 (+31%)
ZFS RAID Z1 223 182 (-18%) 224 ( 0%) 528 (+137%) 189 (-15%) 222 ( 0%)
SYNO SHR-1 234 250 (+ 7%) 264 (+13%) 251 (+ 7%) 266 (+14%) 276 (+18%)
QNAP HW RAID 5 211 231 (+ 9%) 249 (+18%) 226 (+ 7%) 261 (+24%) 273 (+29%)
UnRAID XFS 199 214 (+ 8%) 232 (+17%) 200 (+ 1%) 214 (+ 8%) 242 (+22%)
SR EXT4 BEFORE FILL 86 87 (+ 1%) 87 (+ 1%) 109 (+27%) 86 ( 0%) 85 (- 1%)
SR EXT4 AFTER FILL 88 90 (+ 2%) 90 (+ 2%) 160 (+82%) 97 (+10%) 123 (+40%)
**** COMPARING SATA CONTROLLERS REBUILD TIMES (TIME IN MINUTES - Lower is better):
CMR CMR CMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX
------------- ------------- ------------- ------------- ----------- -----------
LSI 9211-8i SAS 205 226 242 205 253 268
Marvell 9215 SATA 276 302 295 312 314
Onboard Intel H77 216 231 254 219 262 280
**** REBUILD 4x ST2000DM001 CMR to 4x WD20EFAX SMR RAID 5 one disk at a time:
Time in minutes (lower is better)
DISK 1 DISK 2 DISK 3 DISK 4
------------- ------------- ------------- -------------
CMR to SMR MDADM 254 263 258 259
SMR to CMR MDADM 313 302 297 207
CMR to SMR QNAP 260 264 265 266
SMR to CMR QNAP 302 294 266 214
CMR to SMR SYNOLOGY 254 250 291 300
SMR to CMR SYNOLOGY 289 285 282 283
**** SNAPRAID REBUILD FROM PARITY
Results show time in minutes (lower is better)
SYNC 1900GB = Time to build parity 1900GB
SYNC NEW 500GB = Delete 500GB data from data disk, sync parity, write another 500GB random data,
sync back to parity. Last sync time is recorded below
CMR CMR CMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX
------------- ------------- ------------- ------------- ----------- -----------
SYNC 1900GB 191 205 (- 7%) 226 (-18%) 192 (- 1%) 231 (-21%) 245 (-28%)
SYNC NEW 500GB 73 80 (-10%) 74 (- 1%) 72 (+ 1%) 88 (-21%) 95 (-30%)
**** WRITE 1GB FILES AFTER 10MB FILL AND DELETE EVERY OTHER 10MB FILE (MB/SEC):
+XX% = % More speed than ST2000DM001 (better)
-XX% = % Less speed than ST2000DM001 (worse)
1GB WRITE AFTER 10MB FILL AND DELETE EVERY OTHER 10MB FILE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 109 108 (- 1%) 108 (- 1%) 25 (-77%) 65 (-40%) 68 (-38%) 67 (-39%)
ZFS RAID Z1 111 111 ( 0%) 111 ( 0%) 32 (-71%) 98 (-12%) 52 (-53%) 42 (-62%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 40 (-64%) 79 (-29%) 76 (-32%) 76 (-32%)
QNAP HW R5 NTFS 60 59 (- 2%) 56 (- 7%) 27 (-55%) 52 (-13%) 51 (-15%) 67 (-12%)
UnRAID XFS 108 102 (- 6%) 102 (- 2%) 97 (-10%) 43 (-60%) 44 (-59%) 44 (-59%)
SINGLE DISK EXT4 109 102 (- 6%) 102 (- 6%) 11 (-90%) 27 (-75%) 24 (-78%) N/A
SINGLE DISK NTFS 117 83 (-29%) 105 (-10%) 16 (-86%) 36 (-69%) 37 (-68%) N/A
**** READ 1GB FILES AFTER 10MB FILL AND DELETE EVERY OTHER 10MB FILE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 112 112 ( 0%) 111 (- 1%) 109 (+ 3%) 112 ( 0%) 112 ( 0%) 112 ( 0%)
ZFS RAID Z1 109 110 (+ 1%) 109 ( 0%) 97 (-11%) 109 ( 0%) 76 (-30%) 52 (-52%)
SYNO SHR-1 BTRFS 112 111 (- 1%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 111 (- 1%) 111 (- 1%)
QNAP HW R5 NTFS 179 176 (- 2%) 162 (- 9%) 164 (- 8%) 147 (-18%) 143 (-20%) 179 ( 0%)
UnRAID XFS 109 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) 105 (- 4%)
SINGLE DISK EXT4 109 104 (- 5%) 89 (-18%) 90 (-16%) 83 (-24%) 80 (-27%) N/A
SINGLE DISK NTFS 115 104 (-10%) 98 (-15%) 112 (- 3%) 90 (-22%) 92 (-20%) N/A
====================================================================================
#### MIXED FILE WRITE AFTER FULL FILL AND DELETE (MB/SEC)
+XX% = % More speed than ST2000DM001 (better)
-XX% = % Less speed than ST2000DM001 (worse)
**** 10GB WRITE AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 111 111 ( 0%) 111 ( 0%) 105 (- 5%) 110 (- 1%) 106 (- 5%) 106 (- 5%)
ZFS RAID Z1 112 112 ( 0%) 112 ( 0%) 110 (- 2%) 112 ( 0%) 83 (-26%) 112 ( 0%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 110 (- 2%) 112 ( 0%) 112 ( 0%) 112 ( 0%)
QNAP HW R5 NTFS 192 186 (- 3%) 190 (- 1%) 168 (-13%) 191 (- 1%) 194 (+ 1%) 201 (+ 5%)
UnRAID XFS DATA 111 111 ( 0%) 111 ( 0%) 111 ( 0%) 111 ( 0%) 111 ( 0%) N/A
UnRAID XFS PARITY 110 108 (- 2%) 108 (- 2%) 109 (- 1%) 109 (- 1%) 108 (- 2%) N/A
SINGLE DISK EXT4 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) N/A
SINGLE DISK NTFS 154 117 (-24%) 135 (-12%) 80 (-48%) 129 (-16%) 140 (- 9%) N/A
**** 1GB WRITE AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 111 111 ( 0%) 111 ( 0%) 106 (- 5%) 110 (- 1%) 106 (- 5%) 106 (- 5%)
ZFS RAID Z1 111 111 ( 0%) 111 ( 0%) 99 (-11%) 111 ( 0%) 67 (-34%) 112 (+ 1%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 82 (-27%) 112 ( 0%) 112 ( 0%) 112 ( 0%)
QNAP HW R5 NTFS 178 164 (- 8%) 169 (- 1%) 129 (-28%) 170 (- 4%) 194 (- 6%) 179 (+ 1%)
UnRAID XFS DATA 111 110 (- 1%) 111 ( 0%) 106 (- 5%) 110 (- 1%) 110 (- 1%) N/A
UnRAID XFS PARITY 108 106 (- 2%) 105 (- 3%) 110 (+ 2%) 109 (+ 1%) 104 (- 4%) N/A
SINGLE DISK EXT4 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 111 (- 1%) 112 ( 0%) N/A
SINGLE DISK NTFS 134 95 (-29%) 122 (- 9%) 47 (-65%) 115 (-14%) 124 (- 7%) N/A
**** 100MB WRITE AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 106 106 ( 0%) 107 (+ 1%) 19 (-82%) 106 ( 0%) 102 (- 4%) 100 (- 6%)
ZFS RAID Z1 109 109 ( 0%) 109 ( 0%) 97 (-11%) 109 ( 0%) 72 (-34%) 111 (+ 2%)
SYNO SHR-1 BTRFS 109 109 ( 0%) 109 ( 0%) 52 (-52%) 109 ( 0%) 110 (+ 1%) 109 ( 0%)
QNAP HW R5 NTFS 140 122 (-13%) 130 (- 7%) 87 (-28%) 134 (- 4%) 119 (-15%) 141 (+ 1%)
UnRAID XFS DATA 107 106 (- 1%) 98 (- 8%) 106 (- 1%) 91 (-15%) 106 (- 1%) N/A
UnRAID XFS PARITY 104 101 (- 3%) 101 (- 3%) 103 (- 1%) 100 (- 4%) 100 (- 4%) N/A
SINGLE DISK EXT4 109 109 ( 0%) 109 ( 0%) 30 (-72%) 109 ( 0%) 91 (-17%) N/A
SINGLE DISK NTFS 155 109 (-30%) 134 (-14%) 67 (-57%) 137 (-12%) 146 (- 6%) N/A
**** 1MB WRITE AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 45 45 ( 0%) 45 ( 0%) 40 (-11%) 44 (- 2%) 41 (- 9%) 35 (-22%)
ZFS RAID Z1 43 43 ( 0%) 43 ( 0%) 43 ( 0%) 43 ( 0%) 43 ( 0%) 48 (+12%)
SYNO SHR-1 BTRFS 61 60 (- 2%) 61 ( 0%) 58 (- 5%) 60 (- 2%) 70 (+15%) 60 (- 2%)
QNAP HW R5 NTFS 14 12 (-14%) 13 (- 7%) 7 (-50%) 10 (-29%) 10 (-29%) 13 (- 7%)
UnRAID XFS DATA 15 15 ( 0%) 15 ( 0%) 15 ( 0%) 15 ( 0%) 15 ( 0%) N/A
UnRAID XFS PARITY 15 15 ( 0%) 15 ( 0%) 15 ( 0%) 15 ( 0%) 15 ( 0%) N/A
SINGLE DISK EXT4 39 39 ( 0%) 39 ( 0%) 39 ( 0%) 39 ( 0%) 40 (+ 3%) N/A
SINGLE DISK NTFS 259 251 (- 3%) 251 (- 3%) 171 (-34%) 241 (- 7%) 239 (- 8%) N/A
============
#### MIXED FILE READ AFTER FULL FILL AND DELETE (MB/SEC)
+XX% = % More speed than ST2000DM001 (better)
-XX% = % Less speed than ST2000DM001 (worse)
**** 10GB READ AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 109 (- 3%) 105 (- 6%)
ZFS RAID Z1 112 112 ( 0%) 112 ( 0%) 110 (- 2%) 112 ( 0%) 83 (-26%) 72 (-36%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%)
QNAP HW R5 NTFS 234 232 (- 1%) 234 ( 0%) 222 (- 5%) 232 (- 1%) 233 (- 1%) 234 ( 0%)
UnRAID XFS DATA N/A N/A N/A N/A N/A N/A N/A
UnRAID XFS PARITY 109 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) N/A
SINGLE DISK EXT4 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) N/A
SINGLE DISK NTFS 245 236 (- 4%) 215 (-12%) 223 (- 9%) 192 (-22%) 117 (-11%) N/A
**** 1GB READ AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 112 112 ( 0%) 112 ( 0%) 112 ( 0%) 112 ( 0%) 106 (- 5%) 101 (-10%)
ZFS RAID Z1 111 111 ( 0%) 111 ( 0%) 99 (-11%) 111 ( 0%) 67 (-40%) 76 (-31%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 105 (- 6%) 112 ( 0%) 112 ( 0%) 112 ( 0%)
QNAP HW R5 NTFS 230 226 (- 2%) 229 ( 0%) 199 (-13%) 229 ( 0%) 228 (- 1%) 229 ( 0%)
UnRAID XFS DATA N/A N/A N/A N/A N/A N/A N/A
UnRAID XFS PARITY 110 110 ( 0%) 110 ( 0%) 111 (+ 1%) 111 (+ 1%) 110 ( 0%) N/A
SINGLE DISK EXT4 112 112 ( 0%) 112 ( 0%) 108 (- 4%) 112 ( 0%) 112 ( 0%) N/A
SINGLE DISK NTFS 169 156 (- 8%) 147 (-13%) 168 (- 1%) 122 (-27%) 130 (-11%) N/A
**** 100MB READ AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 106 107 (+ 1%) 106 ( 0%) 105 (- 1%) 106 ( 0%) 97 (- 8%) 90 (-15%)
ZFS RAID Z1 109 109 ( 0%) 109 ( 0%) 97 (-11%) 109 ( 0%) 72 (-34%) 73 (-29%)
SYNO SHR-1 BTRFS 112 112 ( 0%) 112 ( 0%) 110 (- 2%) 112 ( 0%) 111 (- 1%) 110 (- 2%)
QNAP HW R5 NTFS 227 224 (- 1%) 226 ( 0%) 160 (-30%) 225 (- 1%) 223 (- 2%) 224 (- 1%)
UnRAID XFS DATA N/A N/A N/A N/A N/A N/A N/A
UnRAID XFS PARITY 109 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) 109 ( 0%) N/A
SINGLE DISK EXT4 110 110 ( 0%) 110 ( 0%) 106 (- 4%) 108 (- 2%) 90 (-18%) N/A
SINGLE DISK NTFS 187 176 (- 6%) 147 (-21%) 198 (+ 6%) 137 (-27%) 144 (-23%) N/A
**** 1MB READ AFTER FILL AND DELETE (MB/SEC):
CMR CMR CMR SMR SMR SMR SMR
SG BARRACUDA WD RED PLUS SG SKYHAWK SG BARRACUDA WD BLUE WD RED 4x WD RED
ST2000DM001 WD20EFZX ST2000VX008 ST2000DM008 WD20EZAZ WD20EFAX WD20EFAX
------------- ------------- ------------- ------------- ----------- ----------- -----------
MDADM RAID 5 EXT4 39 37 (- 5%) 37 (- 5%) 36 (- 8%) 36 (- 8%) 24 (-38%) 15 (-62%)
ZFS RAID Z1 43 43 ( 0%) 43 ( 0%) 43 ( 0%) 43 ( 0%) 43 ( 0%) 19 (-24%)
SYNO SHR-1 BTRFS 74 74 ( 0%) 74 ( 0%) 62 (-16%) 71 (- 4%) 70 (- 5%) 64 (-14%)
QNAP HW R5 NTFS 162 165 (+ 2%) 169 (+ 4%) 85 (-48%) 167 (+ 3%) 99 (-39%) 167 (+ 3%)
UnRAID XFS DATA N/A N/A N/A N/A N/A N/A N/A
UnRAID XFS PARITY 39 39 ( 0%) 39 ( 0%) 39 ( 0%) 39 ( 0%) 39 ( 0%) N/A
SINGLE DISK EXT4 42 40 (- 5%) 38 (-10%) 39 (- 7%) 32 (-24%) 37 (-12%) N/A
SINGLE DISK NTFS 245 236 (- 4%) 215 (-12%) 223 (- 9%) 192 (-22%) 217 (-11%) N/A
====================================================================================
#### TRIM TEST WRITE:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
NOTES: - ST2000DM008 WOULD NOT TRIM IN LINUX (FSTRIM -AV OR DISCARD IN FSTAB)
- USED 3X WD20EZAZ FOR TRIM BECAUSE ONLY ONE VARIATION MODEL OF WD20EZAZ WOULD TRIM IN LINUX, OTHER WOULD NOT
- ALL TRANSFERS OVER 1GBE NETWORK EXCEPT NTFS OVER SATA III
- % = % IMPROVEMENT OF TRIM VS NO TRIM. + VALUES = BETTER, - VALUES = WORSE
**** WRITE 1GB FILES AFTER 10MB FILL AND DELETE EVERY OTHER 10MB FILE (MB/SEC) W/TRIM:
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 68 47 | 47 (+ 0%)
ZFS RAID Z1 52 60 | 90 (+ 33%)
SINGLE DISK EXT4 11 | WNR 27 | 60 (+122%) 24 | 39 (+ 62%)
SINGLE DISK NTFS 16 | 16 (+ 0%) 36 | 130 (+261%) 37 | 85 (+129%)
**** READ 1GB FILES AFTER 10MB FILL AND DELETE EVERY OTHER 10MB FILE (MB/SEC) W/TRIM:
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 112 99 | 96 (- 3%)
ZFS RAID Z1 76 107 | 108 (+ 1%)
SINGLE DISK EXT4 N/A 83 | 92 (+ 11%) 80 | 93 (+ 16%)
SINGLE DISK NTFS 112 | 115 (+ 3%) 90 | 119 (+ 32%) 92 | 95 (+ 3%)
============
**** WRITE PERFORMANCE AFTER FILL AND DELETE (MB/SEC) W/TRIM:
**** 10GB **WRITE** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 106 110 | 110 (+ 0%)
ZFS RAID Z1 83 112 | 112 (+ 0%)
SINGLE DISK EXT4 112 | WNR (N/A %) 112 | 112 (+ 0%) 112 | 112 (+ 0%)
SINGLE DISK NTFS 80 | 127 (+ 59%) 129 | 152 (+ 18%) 140 | 139 (+ 0%)
**** 1GB **WRITE** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 106 109 | 109 (+ 0%)
ZFS RAID Z1 67 112 | 112 (+ 0%)
SINGLE DISK EXT4 112 | WNR (N/A %) 111 | 111 (+ 0%) 112 | 112 (+ 0%)
SINGLE DISK NTFS 47 | 92 (+ 96%) 115 | 138 (+ 20%) 124 | 123 (+ 0%)
**** 100MB **WRITE** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 102 107 | 107 (+ 0%)
ZFS RAID Z1 72 111 | 111 (+ 0%)
SINGLE DISK EXT4 30 | WNR (N/A %) 109 | 109 (+ 0%) 91 | 91 (+ 0%)
SINGLE DISK NTFS 67 | 108 (+ 61%) 137 | 152 (+ 11%) 146 | 143 (- 2%)
**** 1MB **WRITE** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 41 42 | 42 (+ 0%)
ZFS RAID Z1 43 48 | 48 (+ 0%)
SINGLE DISK EXT4 39 | WNR (N/A %) 39 | 39 (+ 0%) 40 | 40 (+ 0%)
SINGLE DISK NTFS 171 | 141 (- 18%) 241 | 313?(+ 30%) 239 | 234 (- 2%)
============
#### TRIM TEST READ:
**** READ PERFORMANCE AFTER FILL AND DELETE (MB/SEC) W/TRIM:
NOTES: - ST2000DM008 WOULD NOT TRIM IN LINUX (FSTRIM -AV OR DISCARD IN FSTAB)
- USED 3X WD20EZAZ FOR TRIM BECAUSE ONLY ONE VARIATION MODEL OF WD20EZAZ WOULD TRIM IN LINUX, OTHER WOULD NOT
- ALL TRANSFERS OVER 1GBE NETWORK EXCEPT NTFS OVER SATA III
- % = % IMPROVEMENT OF TRIM VS NO TRIM. + VALUES = BETTER, - VALUES = WORSE
**** 10GB **READ** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 109 110 | 110 (+ 0%)
ZFS RAID Z1 81 112 | 112 (+ 0%)
SINGLE DISK EXT4 112 | WNR (N/A %) 112 | 112 (+ 0%) 112 | 112 (+ 0%)
SINGLE DISK NTFS 191 | 167 (- 12%) 143 | 165 (+ 15%) 147 | 145 (- 1%)
**** 1GB **READ** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 106 109 | 109 (+ 0%)
ZFS RAID Z1 79 111 | 112 (+ 0%)
SINGLE DISK EXT4 108 | WNR (N/A %) 112 | 112 (+ 0%) 112 | 112 (+ 0%)
SINGLE DISK NTFS 168 | 197 (+ 17%) 122 | 145 (+ 19%) 130 | 128 (- 1%)
**** 100MB **READ** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 97 107 | 107 (+ 0%)
ZFS RAID Z1 75 99 | 100 (+ 1%)
SINGLE DISK EXT4 106 | WNR (N/A %) 108 | 108 (+ 0%) 91 | 90 (- 1%)
SINGLE DISK NTFS 198 | 217 (+ 10%) 137 | 159 (+ 16%) 144 | 146 (+ 1%)
**** 1MB **READ** AFTER FILL AND DELETE (MB/SEC) W/TRIM:
MANUAL TRIM / DISCARD W/4HRS IDLE TIME - AFTER DISK FILL / WRITE (MB/SEC):
SMR SMR SMR SMR
SG BARRACUDA WD BLUE WD RED 3x WD RED
ST2000DM008 WD20EZAZ WD20EZAZ WD20EZAZ
NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM NOTRIM | TRIM
------------- ------------- ------------- -------------
MDADM RAID 5 EXT4 24 42 | 42 (+ 0%)
ZFS RAID Z1 14 18 | 23 (+ 28%)
SINGLE DISK EXT4 39 | WNR (N/A %) 32 | 32 (+ 0%) 37 | 40 (+ 8%)
SINGLE DISK NTFS 223 | 249 (+ 12%) 192 | 284 (+ 48%) 217 | 187 (- 14%)