Plex / QNAP TS-1277 / SSD for PMS/ Docker or not?

[Jake22]

Hello fellow Plexians, I need your advice please on how i should setup my new upgrade from a Synology ds1815+ to a way beefier Qnap TS-1277. Yay! No more random stutters and more space

My usecase:

• I want a setup and forget setup. Plex / Sonarr / Radarr / Jackett / NZBGet / Deluge/
• Flawless and fast Plex experience, quick download and unpack.
• I have 1 nvidia shield, 1 hacked cromebook & 4K TVs - no users outside home.
• Just 1GbE Cat 6 inhouse. 1000/1000 Mbit internet.

Hardware 1:

• QNAP TS-1277-1700-16G High-Performance 12 Bay (8+4) -
• 1 x WD CSSD Red 500GB 2.5 SATA (WDS500G1R0B)
• 4 x Western Digital Red WD100EFAX 10TB (will buy more later on)

Hardware 2:

• Synology ds1815+
• 8 x Western Digital Red 4TB

What i need help figuring out is how to utilize SSD slots on QNAP for PMS and does it make any sense for my use case?

1. Fresh install of QTS to SSD (and buy one more for mirroring. I´d get faster PMS and QTS performance, but i’d be stuck to only download files of up to about 400GB, since my SSD is only 500 GB. Main arry is then installed, for cold storage. With less wear and tear on my expensive 10TB drives.

2. I could use Qtier on QTS 4.3.6, and make one volume for ssd and hdd, but trawling forum posts people don’t seem too thrilled about it. Then again my use case is quite different from most use cases i see on these high end NAS models.

3. I could use SSD as a cache disk?

4. Just drop the SSD/m2 route.

Lastly i’m not quite sure if i should just use QTS packages of Plex / Sonarr / Radarr / Jackett / NZBGet / Deluge/ OR if i should do go the Docker route which seems like all the rage. For me it looks complex and not a setup and forget setup.

Q. If someone has a setup and forget Docker setup please point me in the right direction.

[Coxeroni]

Being new in the QNAP club I can at least elaborate on your last question.

If you want to use docker, you cannot use hardware transcoding as this is limited to the internal apps. Don’t know why, it is just the way it is. The QTS package should be updated by hand by downloading it from here, since the package supplied by QNAP might be behind in terms of version if I recall correctly.

The SSD cache should give you improved read/write speeds, but I don’t think Plex will dramatically benefit from it, since normally drive speed is not the bottle neck. I don’t know if it helps for your file sharing/unpacking business. Same is probably true for Qtier.
@ChuckPa is the QNAP expert probably, he might be able to add his expertise.

[ChuckPa]

If you wish to hear my opinion (most cringe – haha) and how I am setup;

Don’t use Docker (especially on a QNAP) The native app is better and far less fuss on QNAP. Docker may be all the rage but it’s now being abused. It’s intended for use when:

1. A native package doesn’t yet exist for the host platform;
2. For those who frequently move computational load (without data) between hosts. I don’t know of anyone who sets up PMS and then moves it around except for those who use the QTS “move to” feature to actually move a native app from HDD -> SSD volumes.

PMS doesn’t fit any of those criteria.

1. The native exists and It has a ton of data. Add to this:
   a. Hardware transcoding is not native to the container and will never be. You’ll have to add it manually
   b. You can’t move PMS around in the docker container via the GUI. The location is fixed and predetermined by QTS.

   There are those who will demonstrate it can be moved and I’m wrong but by doing so, rule #1 of “Setup and forget” is violated.

   The solution is to use the native apps. Create an extra share or two to be used as if containers for shared data between those apps for common utilities (post processing scripts, mkvmerge, etc). I specifically created usenet share which all the apps point to; /share/usenet/bin and /share/usenet/scripts. These contain the other QNAP tools (mkvmerge, etc)

2. Don’t use QNAP’s “QTier”. If you setup a cache, do so manually which you control. This way, you can turn it off should you ever need to change SSDs (which you will at some point).

Now, for the hardware and its configuration:

1. The TVS-1277, while a good machine, doesn’t have hardware transcoding in the CPU. It may but until Engineering commits to doing so — It may be a very long wait. Until that time, if you want hardware transcoding, you’ll need
   a. The bigger power supply in the QNAP ($$) . Factory is only 250w.
   b. nVidia GPU card (which will use up your full length- 8x PCIe slot)

2. I obtained the TVS-1282 i7-7700 (special order from Taiwan) instead of the TVS-1282 i7-6700 (US stock). The difference is SkyLake (-6700) only supports HEVC SDR in hardware whereas the KabyLake (-7700) supports HEVC HDR (aka UHD) in hardware. It was well worth the wait.

My configuration:

Hardware:

1. DS1815+ (8x 8TB WD Red Pro) – offline monthly backup for the QNAP and Synology development / testing platform.

2. TVS-1282-i7-7700-32GB w/
   a. 2x 1TB Samsung 860 EVO SATA-3 M.2 SSD - CACHEDEV1 - QTS main (RAID 1)
   b. 8x 8TB WD Red Pro - CACHEDEV2 - Main data array
   c. 1x 500GB Samsung 860 EVO 2.5" SSD - Incoming media processing volume (Sonarr, etc) which then transfers to main array when done.
   d. 2x 500GB Samsung 860 EVO 2.5" SSD - VMWare Workstation VMs (RAID 1)
   e. 1x 500GB Samsung 860 EVO 2.5" SSD - Plex database & other development
   f. 1x QM2-2P-384 PCI-E NVMe SSD carrier / interface.
   g. 1x Samsung 970 Pro NVMe 1TB – R/W cache for main volume (CACHEDEV2)
   h. QNAP (Aquantia AQC 107) 10 GbE - single port NIC

3. Netgear - GS110emx - 10 port (8x 1GbE, 2x 10 GbE / Mgig) managed switch
   Software config:

4. Intel NUC8-i7HVK w/ Aktio 10 GbE Thunderbolt 10 GbE adapter - Workstation

5. 2x ASUS MG280UQ 2160p monitors as joined desktop.

Performance:

1. Sonarr & NZBget were setup and forget.
2. PMS is setup & forget (except when I break it doing development)
3. 10 GbE and cache sustain full wire speed (1 GB/sec) read / write
4. i7-7700 has demonstrated transcoding 6 simultaneous 2160p HEVC 80 MB/s -> 1080p H.264. I ran out of clients in the house to test on. CPU load was 65%

Secondary:

1. Secondary media processing and other development work on the NUC sees the QNAP shares as faster-than-local HDDs.
2. No perceived difference between VMs on local NVMe (NUC) vs NFS share.

Final:

This is overkill for most. It is a development and testing configuration; a media ingestion configuration; and lastly a whole house media server.

My recommendation, if finances allow:

a. TVS-1282 instead of TVS-1277 (price is the same before adding bigger supply)
b. Skip the nVidia GPU - extra cost in both board and power supply option in QNAP.
c. Use the native apps. Avoid Docker. Why use abstraction when it limits options.
d. nVidia GPU HW transcoding is inoperative on QNAP until QTS 4.4.2. I have been working with QNAP to develop, and perform final testing of, the updated NVIDIA_GPU_DRV package which is required for PMS 1.17.0 and above.

Coming soon to QTS:

I am about to open Forum Preview Testing for new QNAP packaging. In that packaging. The new packaging follows the architecture used on Synology (Package Center & Plex share)

1. Uninstalling the app will save your installation metadata prior to removal (docker won’t). Accidentally uninstalling will no longer be data-fatal.
2. Installing the app will recover (pull from an admin-only share) your previously saved metadata.
3. With the package uninstalled, you can backup and restore previous backup versions of your metadata

I will offer any additional help I can. Don’t be afraid to ask. (I don’t bite – too hard)

[Jake22]

Firstly a big thank you, for your well rounded answers.

I will not go the docker route then, and just use the native apps to QTS.

I already bought the TVS-1277, sorry for not being crystal clear about that. I got a really good deal on the TVS-1277, so it’s not an option to switch.
The Power Supply Unit is - “550W, 100-240V AC, 50/60 Hz, with additional graphics card power cables”, so it should be possible to add a Nvidia GFX in the future if i need it, and its supported. But i hope it can software transcode enough for my needs, which is basically one screen - set to passthrough. It’s my understanding that PMS transcodes if there are subtitles present.

I will follow your recommendation of not using Qtier, but it’s a bit unclear to me what you recommend i should do with SSD in my setup.

“Don’t use QNAP’s “QTier”. If you setup a cache, do so manually which you control. This way, you can turn it off should you ever need to change SSDs (which you will at some point).”

“a. 2x 1TB Samsung 860 EVO SATA-3 M.2 SSD - CACHEDEV1 - QTS main (RAID 1)

b. 8x 8TB WD Red Pro - CACHEDEV2 - Main data array

c. 1x 500GB Samsung 860 EVO 2.5" SSD - Incoming media processing volume (Sonarr, etc) which then transfers to main array when done.

e. 1x 500GB Samsung 860 EVO 2.5" SSD - Plex database & other development”

I’ve only bought one Red 500GB 2.5 SSD, but i can get a few more if it makes any sense for the speed of the setup. If not i’ll just save the money, and wait for SSD´s to become even cheaper

1. Would you recommend i get more SSD´s or m2 for QTS (RAID 1).
2. Install QTS on my main raid 4x 8TB WD Red Pro raid6 (will be expanded later on), and use my single Red 500GB 2.5 SSD in JBOD for Incoming media processing volume (Sonarr, etc) which then transfers to main array when done.
3. Use SSD as a cache drive, even though its much smaller than the main raid. Ive heard there is a 10% rule SSD/HDD.
4. ?

Ps. It’s good to bite once in a while

[ChuckPa]

Thanks for the clarification on the TVS-1277 w/ 550w supply.
You’re all set for power. Good deal. I contemplated getting the supply and upgrading myself — for all of about 5 seconds!

Let me show you my caching:

Hardware:

Config:

If you want the 500 GB for media processing in your 2.5" slot. that’s fine. It’s actually a good idea to not run that through the cache (why waste the SSD?)

1. Install QTS on the internal M.2 SSD RAID 1 array. It’s blazing fast as you see below.

2. The main array (CACHEDEV2) is setup as primary storage (you create the array after creating / installing QTS as RAID 1 on the M.2 internal SATA-3 SSDs.)

3. Initialize & assign the PCIe SSD as cache for CACHEDEV2. It makes no sense to use SATA-3 (at 535 MB/sec) in the SSD when the main array is capable of 1.0 GB/sec. This is why I use the PCIe SSD.

Notice here the FS_Throughput

1. QTS volume at 900+ MB/sec (Read = 1.06G, Write = 535; read is striped)
2. Main volume HDs sustain 700MB. PCIe cache boosts that to 3.05 GB
3. DataVol3 = 2.5" media processing SSD (I had a spinner but it was way too slow)
4. DataVol4 = VM SSD RAID

[~] # qcli_storage -tT
fio test command for LV layer: /sbin/fio --filename=test_device --direct=0 --rw=read --bs=1M --runtime=15 --name=test-read --ioengine=libaio --iodepth=32 &>/tmp/qcli_storage.log
fio test command for File system: /sbin/fio --filename=test_device/qcli_storage --direct=0 --rw=read --bs=1M --runtime=15 --name=test-read --ioengine=libaio --iodepth=32 --size=128m &>/tmp/qcli_storage.log
Start testing!
Performance test is finished 100.000%...
VolID   VolName             Pool     Mapping_Name            Throughput      Mount_Path                    FS_Throughput
1       DataVol1            288      /dev/mapper/cachedev1   1.04 GB/s       /share/CACHEDEV1_DATA         941.18 MB/s
2       DataVol2            289      /dev/mapper/cachedev2   700.48 MB/s     /share/CACHEDEV2_DATA         3.05 GB/s 
3       DataVol3            290      /dev/mapper/cachedev3   537.23 MB/s     /share/CACHEDEV3_DATA         533.33 MB/s
4       DataVol4            291      /dev/mapper/cachedev4   1.02 GB/s       /share/CACHEDEV4_DATA         1.01 GB/s 
[~] # 

As for the SSD vs HDD prices, if I may?

https://www.samsung.com/us/computing/memory-storage/solid-state-drives

Go to the full specs (at the bottom)
Look at the TBW for each size. Do the math. You’ll discover the larger drives are actually much cheaper in the long haul. Greater TBW because there are more SSD pages which can be used. Less expensive per GB.