120 HDD on single PSU

JBOD is not RAID

JBOD = Just a Bunch of Disks.

That is not true. Take a look at “What do Bronze, SIlver, … ratings mean?” section, and check values for Titanium certification. It is up to 94% at 100% load. Again, PSU with one sticker is not the same as with another. Also, from brand A is not the same as from brand B. As we already stated some time ago, saving on PSU quality is potentially the worst one can do, overestimating what it can deliver (thus underestimating the load) is another bad thing to do.

Kudos to Voodoo for finding a real world answer. I thought the most relevant quote form Voodoo’s link to the question that stated this thread is below:

So AlexXXL With enderTown’s example, I would say that you could barely power your 120 HDDs with two 1200 watt server quality PSUs. I would go with two 1500 watt server quality PSUs to be safe and solid. This also assumes that you a safe wiring setup
If you do go with another 100 HDDs I would add two more 1500 watt PSUs and spread the power equally.

That describes efficiency ratings and has nothing to do with whether a device performs well running at 100% constantly.

"What percentage should PSU run at?

Moreover, you should keep in mind that the sweet efficiency spot is usually around 40-50% of the PSU’s max-rated capacity. That and the closest a given power supply is to its max load output, the lowest the efficiency.Apr. 21, 2021"

or this

or this

https://pcpartpicker.com/forums/topic/77606-how-much-extra-wattage-should-a-psu-have

and many other articles, guides, and even manufacturer recommendations will say the same.

High End power supplies will run safely, full time, at their max rated PSU but it WILL shorten their life doing so. Low end PSUs can literally catch on fire if left on 24/7/365 at full load with maybe a bit too much dust or a weak solder connection.

Most recommend that if you do not want to shorten the life of your PSU you should run full time at no more than 50-70% capacity and any PSU is more subject to failure and/or fire if run at max capacity constantly.

With a server grade PSU I would run full time at or near 100% capacity if I were not concerned about its lifetime but I would prefer to use under 70%. In any case of a full time hard working PSU I would take care to keep my PSU clean, cool, and regularly inspected.

We don’t have to argue, but putting 120 HDD on a PSU will most certainly blow the +12v and +5v rails. That is a fact. The 80+ ratings are only efficiency guidelines at continuous load. It doesn’t say anything about burst loads. Besides, if, for example, you put 1000w of load on a 1200w PSU, burst loads will almost certainly exceed 1200w and destroy the PSU.

2 Likes

how do you assign drive letters with that many drives. I’ve been using NAS in raid to limit my drive letter usage. thanks matthewjbauer. Found this link if anyone is interested . How to mount hard drive as folder on Windows 10 | Windows Central

2 Likes

You don’t. At least in Windows, you create an NTFS folder and mount the logical folder as a physical drive.

2 Likes

Thank you. I appreciate you insight into my problem.

1 Like

It depends on how those disks are connected to your controller:

From wikipedia:

There is a pertinent paragraph that reads:
" JBOD[edit]

JBOD (abbreviated from “Just a Bunch Of Disks”/“Just a Bunch Of Drives”) is an architecture using multiple hard drives exposed as individual devices. Hard drives may be treated independently or may be combined into one or more logical volumes using a volume manager like LVM or mdadm, or a device-spanning filesystem like btrfs; such volumes are usually called “spanned” or “linear | SPAN | BIG”.[2][3][4] A spanned volume provides no redundancy, so failure of a single hard drive amounts to failure of the whole logical volume.[5][6] Redundancy for resilience and/or bandwidth improvement may be provided, in software, at a higher level."

The key words are:
“…so failure of a single hard drive amounts to failure of the whole logical volume…”

So it depends on how user rfc2324 set up his JBOD, which is why I commented with the potential risk.

Cheers!

2 Likes

sure, but that goes for any hdd’s you have attached to a computer.
I would say the key phrase of jbod is:
“is an architecture using multiple hard drives exposed as individual devices”

Unless you specifically choose to use Raid(like) software of controller to create a spanned volume, it is just what it says it is, a bunch of disks.

There is no inherent risk of using a Jbod disk rack, there is an inherent risk of using spanned volumes without redundancy,

1 Like

The “risk” depends on the value of what needs to be protected. There is no value in plots, except replotting times. As such for hosting plots (spanned or RAIDed), if one doesn’t have redundancy, replotting time is potentially a long one. On the other hand, with redundancy, the cost of replotting one failed drive is much lower than the revenue lost due to redundancy. Either way, spanned or RAID architecture is not the best solution for hosting plots (IMO).

Good post. Just to make sure, I have the specs of the mentioned PSU:
|+5V|25A|
|+3,3V|25A|
|+5V & +3,3V combined|150W|

|+12V1|104A|
|+12V2|-|
|+12V1, +12V2 combined|1248W|

I will need the specs for the HDDs to calculate the power needed according to the specs. I will have them tomorrow and post here the results. But even without the specs, I can see that I have 150W for 3v+5V, so if the drive needs 3-5W from a 5V source, I have between 30-50 HDDs on this supply.
Plus, I have read recently that on most modern PSUs, 12V and 5V are coming from same “source”, meaning 5V is provided by lowering the V from 12V to 5V. This maybe means that I can be pretty comfortable regarding the 5V power needs, if that is the case with my supply, of course.

Well can we calculate this together?
If normal power usage of a hdd is less than 10W, each molex to 5 x sata adater will have to endure less than 50W, which is quite real. The real problem us the connector on the PSU which by this math will have to endure eather 3x molex to 5 x sata, which means less than 150W or more, if I put cables with more molex to 5 x sata converters, like 4-6 of them.

I have certain experience with GPU mining and can tell you that PSUs can work 24/7 on 90% load or even more in some cases. For months, if not for years.

User “rfc2324” wrote that he intends for his JBOD to consist of 24 disks.
That is a big risk, if set up such that a single failure blows the data on all 24 disks.

Hence, my cautionary comment.

Yes, that is a key phrase. But that does not mean that other key phrases are not key phrases, such as:
“A spanned volume provides no redundancy, so failure of a single hard drive amounts to failure of the whole logical volume.”

Most hardware RAID controllers support JBOD.
There are people that use that feature, or it would not be on nearly every hardware RAID controller.

User “rfc2324” did not state how he set up his JBOD. Ergo, my cautionary note.

I err on the side of caution, and not assumption.

JBOD as already explained above is just a way of how drives are exposed, not how they are handled. They can be either handled as individual drives or as a spanned volume. Therefore, your assertion only applies to handling JBOD drives as a spanned volume.

We agree on this one :slight_smile:

@AlexXXL Take a look at this datasheet:

I understand that those are not your HDs, but it gives some idea how power is used. During the startup, they draw ~22W, but when they settle down, it is only 5.6W. Therefore, the question is whether you can address power-on staggering of those drives.

I would also assume that 5V is used to power the electronics, where 12V is for the spindle and heads. Therefore, you may not need to worry about power draw on 5V line.

Yes, I am aware of all you say and I am calculating those W too. As I mentioned before, most of the HDDs are on LSI RAID cards, so hopefully, I will manage to address power-on staggering. In case I don’t, I can always add one more 1200W PSU. Or in the worst case, I will find a way of powering on the drives in groups, like a group of 20-30 HDDs at once, keeping the power draw in control while spinning up the drives. There must be some kind of power on/of switch I can use for this.

If you need to depend on a manual switch, just make sure that in the BIOS you disable power-on on power loss.