DS1621+ NVMe recommendation

Currently reading
DS1621+ NVMe recommendation

you can disagree, you are entitled to do it
but
it doesn't change the fact that DSM in original setup, w/o tweaking can’t support PLP. Please read carefully all 4 obstacles and think twice about them.

One of the possible change to step in is usage of controlled operation by “ups.status” and ups@localhost described here:

and regarding this your statement:
By the way, Synology also thinks that PLP is usefull in Syno environment, at least they provide PLP in their SNV-3500 NVMe SSD.
... in this case the “useful” wording is just a wish.

Btw. PLP will save few tenths of microseconds, that can help save necessary metadata (mapping table). Then better investment is to sufficient and stable power from UPS and managed environment between NAS and UPS. For all your data and all your disks.

Provided link can help you save investment to:
- pointless PLP (SSD) feature in DSM environment
- your data protection based on insufficient UPS management in DSM
 
you can disagree, you are entitled to do it
but
it doesn't change the fact that DSM in original setup, w/o tweaking can’t support PLP. Please read carefully all 4 obstacles and think twice about them.

One of the possible change to step in is usage of controlled operation by “ups.status” and ups@localhost described here:


and regarding this your statement:

... in this case the “useful” wording is just a wish.

Btw. PLP will save few tenths of microseconds, that can help save necessary metadata (mapping table). Then better investment is to sufficient and stable power from UPS and managed environment between NAS and UPS. For all your data and all your disks.

Provided link can help you save investment to:
- pointless PLP (SSD) feature in DSM environment
- your data protection based on insufficient UPS management in DSM
Thank you for your detailed explanation, jeyare. Then I might be wrong about how does PLP work.
I thought that using a UPS and having PLP equipped NVMe drives are totally different story which are independent from each other.

This is what I thought about using a UPS:
- Using a UPS will enusre that your syno still gets power in case of electricity blackout
- Once the battery status of your UPS is low, your Syno will unmount all the volumes to make sure that there will be no data loss happen once the battery is empty and your NAS will be off

This is what I thought about using a PLP equipped NVMe drives:
- Power Loss Protection will make sure that even in case of an unexpected power loss there will be enough electricity in the capacitor of your NVMe drive to write all the data from the DRAM of the NVMe drive to the storage or the NVMe drive, so there will be no lost data
- I thought it depends only the hardware (and the firmware) of the NVMe drive, no OS support is needed.

As far as I understand I'm wrong with these, am I?
 
Y and N ... there isn’t a simple answer
both are linked, for an understanding you have to dive to deeper levels

1. Syno setup of UPS in DSM:
- two possible triggers of NAS Safe mode switch on:
a) Low battery status, based on UPS firmware settings (out of NAS control). You can setup, in each UPS the level of the “Low” (up to UPS system). Frequently based on % value of rest of capacity.
b) time on battery. This is one of half-baked features in DSM.

Then just thinking:
- you have defined trigger in your UPS for Low state = 20% of the remaining capacity
- your UPS has self defense trigger (switch of) e.g. from 10% (you can’t drain 100% of the capacity = decreasing of the lifespan)
- you don’t have precise calculated UPS characteristics (there isn’t not just Power capacity)
1. You have heavy load in your NAS, then you can drain your batteries faster.
2. But when you have higher load from NAS, then character of battery load curve will provide an answer = you will drain your batteries faster. Physical principles.
Conclusion:
- time of battery draining between 21% and 10% depends on the capacity of the batteries and the load (for small batteries it could be 10 seconds)
- also depends on quality of entire UPS controller
- also depends on lifespan of the batteries. Older batteries you can drain really faster in last 25% capacity (batteries characteristics)
then your UPS can step into self protection and can switch of faster than Syno finished Safe mode (maybe your problem).
OFC, there are low cost UPS w/o self protection.

In any cases you don’t need PLP SSD feature, because one of the commands from DSM in beginning of the Safe mode is ATA controller command= Stanby Immediate Command (abort data write and reflush your buffer = PLP), then DSM will stop all services and then DSM will unmount all volumes. You need invest just into time.

You need just proper UPS HW and SW whats is in 3kW level less than two enterprise SSD with PLP.
Then you need control the shutdown of your NAS based on dynamic change of the defined % trigger, measured from UPS by NAS (guidelines provided).
Then you can operate any kind of disks without PLP, but with better data safety than PLP.

2. right PLP feature operation is dependent from host controller (here is the point of the linked environment). As was mentioned by me in previous post - the mapping table = flash transition layer. And here is a critical point what OS and what FS are used for the operation with the controller = data alignment to transferred size. There is another critical point when the misalignments could cause data fragmentation during transfer across the SATA bus to the SSD and you can lose more time than tenths of microseconds (purchased by small capacitor in the SSD). Then you lose data also.
And think about standard random write operation. The fragmentation is growing with the amount of random I/Os from parallel processes (OS, FS dependencies).
Then you need more than tenths microseconds from PLP. You need minute (and more, up to systems) to be save.

Final conclusion:
I can’t imagine why is purchase of the enterprise SSD include PLP (as one of the driver) aligned with NAS operation without proper care. And proper UPS belongs to enterprise level storage mediums operation. OFC to any kind of disks.
Problem of the DSM is in the proper cooperation with standardized UPS systems w/o tweaking of the DSM by users.
 
you're welcome
I have more scientific approach, rather than Y or N answers. Then some of my posts are long.
because:
"For every complex problem there is an answer that is clear, simple, and wrong."

... more complex description what big difference is between SSD contains:
- simple PLP (capacitor only)
- advanced PLP with onboard independent MCU based power control:
a) Over Input Voltage Protection
b) Power-up Inrush Current Suppression
c) Input Power Noise De-glitch
d) and ofc: Fast Power On-Off Control

1609145662477.png


source from one of the industrial NAND producers - atp.com:

 
My NVMe SSD is an ATP product and is equipped with MCU, the protection which is mentioned in the article you linked. As far as I understand MCU will make sure that this SSD will always be able to reboot (even after a power supply failor, for example). It will also make sure that there will be no data lost within the NVMe SSD. (I understand that still data loss might occur in case of power failor because of no proper support from the OS.) Is that correct?
 
As far as I understand MCU will make sure that this SSD will always be able to reboot (even after a power supply failor, for example). It will also make sure that there will be no data lost within the NVMe SSD. (I understand that still data loss might occur in case of power failor because of no proper support from the OS.) Is that correct?

I think, there is missing reading with an understanding. No one can give you a guarantee that your SSD will ALWAYS be able to reboot.
It just can help you to avoid SOME situations than could be lead to troubles after rebooting. And it is far away from ALWAYS. To be honest I don’t remember for such statements from any disk vendors.

Same for the data loss. You can’t replace right and heavy UPS operation:
- power filtration;
- power stabilization;
- finally +10minutes of the critical time for safe shutdown;
by the MCU with capacitor in SSD. It’s the physics mentioned above.
 
regarding the PLP (Power Loss Protection), for a proper operation of the PLP you need 4 mandatory cooperated parts in your NAS:
1. NAS backed by UPS and correct communication between the two components (not just Power Low info)
2. NAS system (DSM) with ability to know of the certain level of each statuses from UPS
3. Disk controller with ability to provide “ a message” for host to stop transfer data in the right moment (critical feature fir the PLP)
4. NAS system with the ability to read data from NAND back to DRAM within the SSD (check of saved data during supported microseconds).

so and Syno DSM is far away from this kind of scenario support (specially fro first two).
Then PLP feature in SSD is pointless in Syno environment.
Sorry, sir, this is an missconception on what an SSD/NVMe PLP protection is and what protect.

1st PLP is not to protect your data from incomplete system writes corruption, as use to happen with spinners, such thing is duty of the OS's File System Journal

2nd then what is it for? is to enable the SSD complete controller pending write operations, is not the same? No, SSD, NVMEe are made from NAND elements with limited write cycles, the controller keeps tables which along an algorithm enable to write such NAND just enough times before it expires its maximun writes, but this table is only updated after an full write operation is complete, so PLP provides energy to complete such incomplete writes and update the tables before an catastrophic power failure, without such protection, there's no warranty on the NAND condition and may experiment more writes than it is intended for, with unforeseeable consequences.

3rd the process you describe should be mandatory to every NAS w/o PLP, having an UPS with spare energy to complete an safe shutdown, this will ensure NO actual POWER LOSS incident would corrupt your's SSD NAND.

Accordignly, PLP-SSD/NVMe actually are not necesary if you have an GOOD UPS with generous margin and Plan Ahead for an Blackout Event for proper shutdown, here the UPS and SSD enemy are the HEAVY APPLICATIONS, some applications may interfere an shutdown by CPU/Memory monopolization, commonly this may happen if you run an database reindexation in a Multi-GB database (as Bitcoin blockchain), but not the only case, however depends on CPU load and Applications, if you want to run an Bitcoin Node, at least during IBD you should triple UPS capacity or go for PLP NVMe.
 

Create an account or login to comment

You must be a member in order to leave a comment

Create account

Create an account on our community. It's easy!

Log in

Already have an account? Log in here.

Similar threads

  • Question
No, No and No…. Left to itself, ABB will consume data space until there is none. You must create a...
Replies
17
Views
1,411

Welcome to SynoForum.com!

SynoForum.com is an unofficial Synology forum for NAS owners and enthusiasts.

Registration is free, easy and fast!

Back
Top