I put my PMS installation out on the expansion 2.5" slot for several reasons:
- The mechanical aspect (dismantling) versus backing up, dismounting in software when change is needed, copying the PMS library back (aka. “Migrate to”)
- Endurance of the device versus cost.
- Functional loading (tasks) I need it for.
a. Always installing PMS versions and testing them
b. My own packaging development in the VM need a SSD
Mirrored M.2 SSDs, as created by QTS, will wear out at the same time. The same data is written to both simultaneously. The true performance comes when reading. QTS reads in a striped manner. Measured performance: 535 MB/sec write, 1.05 GB/sec read (1 & 2 below)
Package development and PMS testing speed (building new libraries) is limited by my internet (44 Mbps). 535 MB/sec is more than adequate for this task.
[~] # qcli_storage -Tt force=1
fio test command for physical disk: /sbin/fio --filename=test_device --direct=1 --rw=read --bs=1M --runtime=15 --name=test-read --ioengine=libaio --iodepth=32 &>/tmp/qcli_storage.log
fio test command for RAID: /sbin/fio --filename=test_device --direct=0 --rw=read --bs=1M --runtime=15 --name=test-read --ioengine=libaio --iodepth=32 &>/tmp/qcli_storage.log
Start testing!
Performance test is finished 100.000%...
Enclosure Port Sys_Name Throughput RAID RAID_Type RAID_Throughput Pool
NAS_HOST 1 /dev/sda 534.47 MB/s /dev/md1 RAID 1 1.05 GB/s 288
NAS_HOST 2 /dev/sdb 538.58 MB/s /dev/md1 RAID 1 1.05 GB/s 288
NAS_HOST 3 /dev/sde 535.46 MB/s -- -- -- --
NAS_HOST 4 /dev/sdc 133.06 MB/s /dev/md3 Single 138.17 MB/s 290
NAS_HOST 5 /dev/sdf 531.88 MB/s /dev/md4 Single 534.26 MB/s 291
NAS_HOST 6 /dev/sdd 108.57 MB/s /dev/md5 Single 108.86 MB/s 292
NAS_HOST 7 /dev/sdh 201.15 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 8 /dev/sdg 192.99 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 9 /dev/sdn 190.75 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 10 /dev/sdm 194.94 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 11 /dev/sdk 201.91 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 12 /dev/sdl 193.06 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 13 /dev/sdi 186.38 MB/s /dev/md2 RAID 5 1.20 GB/s 289
NAS_HOST 14 /dev/sdj 194.22 MB/s /dev/md2 RAID 5 1.20 GB/s 289
[~] #
The breakdown results are:
- Changing the M.
My TVS-1282 has Samsung 860 1TB M.2 SSDs in it. They are a pain to install.
Because of all the PMS installations (upgrades, downgrades, adding test data) I do for users, my PMS installation is pretty brutally used.
The performance gained (535 vs 1016 MB/sec) is insignificant compared to the time required to transmit the info from PMS to the client (even at 5 Gbps) and the client to render the display. (the slowest link in the chain)
In their QTS-created RAID configuration (mirror), both are written simultaneously therefore both will wear out at the same time. I
Because of how mirroring works, and both being written to simultaneously, but duplicated, the total observed performance is ~535 MB/sec. (speed comes from reading in a striped manner)
Given the write speed of a 2.5" SSD is the same as the M.2 (both SATA-3), there is no net performance gain when building PMS databases (which I do a lot of)
My evaluation:
- Changing a 2.5" SSD in the expansion bay is infinitely easier than pulling out the unit, opening the cover and removing SSDs from the motherboard.
- Cost: 2.5" Samsung 860 EVO 1TB is $129 (B&H Camera), M.2 Samsung 860 EVO 1TB is $169.
- Performance is the same. Both are SATA-3 devices
- Endurance is the same
Final decision for me:
- Ease of replacement 5 minutes vs 1 hr
- Cost of replacing two SSDs: ($169 each) vs $129
- Database read speed difference on single SSD (535 MB/sec) vs CACHEDEV1 RAID SSD (1.05 GB/sec)) is insignificant / noise when compared to the time required to send the data to the player and the player to render that information on the display
- Database write speed is identical as both have identical write speeds.
