Fastest way transfer files onto NAS

[airborne spoon]

My old NAS took a crap and I had to transfer all the files (11tb) to a new spare 18tb drive. But when I did that it was all on one system with a live USB of Ubuntu.

 

I have been building a new NAS with server hardware and it should be operational today once the ECC RAM arrives in the mail.

 

I will be installing trueNAS and using it for Plex. I have a fiber 10GB line between my gaming PC and the server. When I built it years ago using old gaming PC components I just transferred in Windows via drag and drop. It took quite a while and I only had like 5tb at that time.

 

Is there a faster way to transfer that many files either in windows or Ubuntu?

I used rsync to transfer them originally during my recent troubles via terminal and it took about 17 hours.

[Kilrah]

With 10G networking the limit will be the speed of the drive(s) themselves, in this case the 18TB one assuming you'll have an array at the other end.

 

IF you have a way to connect/mount the drive on truenas itself and run the copy locally that could be a little faster but I only know enough about truenas to know I don't like it  

[tkitch]

nope, because a single HDD is your bottleneck for speeds there.

 

 

[Kid.Lazer]

As other's have mentioned, you'll be limited by your single drive. Unless it's something spectacular, a single spinning rust drive will get you ~250MBps which would take about 12 hours (if my napkin math is right.)

[airborne spoon]

8 minutes ago, Kilrah said:

With 10G networking the limit will be the speed of the drive(s) themselves, in this case the 18TB one assuming you'll have an array at the other end.

 

IF you have a way to connect/mount the drive on truenas itself and run the copy locally that could be a little faster but I only know enough about truenas to know I don't like it  

Yeah it's 8x 10tb WD red drives in raidz2

[tkitch]

Just now, airborne spoon said:

Yeah it's 8x 10tb WD red drives in raidz2

a RAID of at least 3 or 4 drives will always be faster than one drive.
the RAID isn't your concern.

That one lone USB HDD is the slow part.

 

[podkall]

8 minutes ago, tkitch said:

a RAID of at least 3 or 4 drives will always be faster than one drive.
the RAID isn't your concern.

That one lone USB HDD is the slow part.

 

does HDD read or write faster?

[Eigenvektor]

3 minutes ago, podkall said:

does HDD read or write faster?

A read should generally be faster, but you potentially have some latency for seeks if data is non-consecutive, which might slow down the transfer.

 

But a RAID can usually distribute both reads and writes across multiple drives at once, which gives it a performance advantage over a single drive.

[Kilrah]

3 minutes ago, podkall said:

does HDD read or write faster?

On modern drives it's typically the same for both. At least for the drive itself, then there are lots of OS-level things that can affect it.

[airborne spoon]

15 minutes ago, tkitch said:

That one lone USB HDD is the slow part.

 

Who said anything about USB? Now that would be slow. Naw it's connected to the motherboard via SATA port in my gaming PC

[Kilrah]

5 minutes ago, airborne spoon said:

Who said anything about USB? Now that would be slow.

Nope, not slower when connected via USB, unless you're somehow using USB2... USB3 is capable of maxing out an HDD just fine.

[airborne spoon]

11 minutes ago, Kilrah said:

Nope, not slower when connected via USB, unless you're somehow using USB2... USB3 is capable of maxing out an HDD just fine.

Interesting well if I had a USB3 carrier I would do that but all I have is an old one from like 2007 that uses a mini-USB plug. So the old run a temp SATA cable to a HDD sitting next to the case is what I'm rolling with lol

[tkitch]

21 minutes ago, airborne spoon said:

Interesting well if I had a USB3 carrier I would do that but all I have is an old one from like 2007 that uses a mini-USB plug. So the old run a temp SATA cable to a HDD sitting next to the case is what I'm rolling with lol

if it's USB 2, you'd be slowing it down dramatically.
Micro USB / Mini USB are both USB 2.

USB 3 can push triple what most HDDs can push, more of it's 10 or 20gbps USB 3. 

 

check the drive connector on the USB Enclosure compared to this:

[airborne spoon]

46 minutes ago, tkitch said:

if it's USB 2, you'd be slowing it down dramatically.
Micro USB / Mini USB are both USB 2.

USB 3 can push triple what most HDDs can push, more of it's 10 or 20gbps USB 3. 

 

check the drive connector on the USB Enclosure compared to this:

Bro I said it's from like 2007 it's ancient

I know good and well the difference between micro, mini and all the USB shapes and sizes. Obviously I wasn't tracking the speeds but that's a moot point.

[podkall]

2 hours ago, Eigenvektor said:

A read should generally be faster, but you potentially have some latency for seeks if data is non-consecutive, which might slow down the transfer.

 

But a RAID can usually distribute both reads and writes across multiple drives at once, which gives it a performance advantage over a single drive.

how does having data on multiple drives save time/improve performance?

[Eigenvektor]

20 minutes ago, podkall said:

how does having data on multiple drives save time/improve performance?

If you have the same data on multiple drives, you can read from them simultaneously.

 

For example, let's say you have two drives. Then you could read bytes at index 0, 2, 4, … from one drive and bytes at index 1, 3, 5, … from the other one at the same time, effectively doubling your read speed. Provided the controller and bus it is connected to doesn't turn into a limit (which would most likely be the case for NVMe)

 

For write speed, it depends. RAID 0 means that you write separate halves of the data to two drives at the same time. It can be combined with other RAID levels (e.g. 5+0). Other levels generally decrease performance because of the parity bits that need to be calculated/written

[podkall]

2 minutes ago, Eigenvektor said:

If you have the same data on multiple drives, you can read from them simultaneously.

 

For example, let's say you have two drives. Then you could read bytes at index 0, 2, 4, … from one drive and bytes at index 1, 3, 5, … from the other one at the same time, effectively doubling your read speed. Provided the controller and bus it is connected to doesn't turn into a limit (which would most likely be the case for NVMe)

 

For write speed, it depends. RAID 0 means that you write separate halves of the data to two drives at the same time. It can be combined with other RAID levels (e.g. 5+0). Other levels generally decrease performance because of the parity bits that need to be calculated/written

does the data write multiple times or is it just across whoever is first available?

[tkitch]

5 minutes ago, podkall said:

does the data write multiple times or is it just across whoever is first available?

depends on the type of RAID you're running.    

(This convo is probably worth a thread of its own)

[podkall]

4 minutes ago, tkitch said:

depends on the type of RAID you're running.    

(This convo is probably worth a thread of its own)

say less

[Kilrah]

41 minutes ago, podkall said:

does the data write multiple times or is it just across whoever is first available?

All depends on the array type and setup, but e.g. if you have a 5-drive raidz1, i.e with protection against failure of one drive, anything you write, plus the redundancy data gets split into equal amounts that go to each of the 5 drives.

 

So in principle instead of writing the data to one drive and be limited by its speed you now write a 5th of the data to 5 drives at the same time, so it's 5x faster.

There's some overhead compared to the raw file for the redundancy data, calculations and synchronization so it'll be a bit less, but you get the idea.

 

This has data for a single drive at 108MB/s writes, and a raidz1 of 5 of the same drive at 469MB/s.

 

https://calomel.org/zfs_raid_speed_capacity.html

 

[podkall]

15 minutes ago, Kilrah said:

All depends on the array type and setup, but e.g. if you have a 5-drive raidz1, i.e with protection against failure of one drive, anything you write, plus the redundancy data gets split into equal amounts that go to each of the 5 drives.

 

So in principle instead of writing the data to one drive and be limited by its speed you now write a 5th of the data to 5 drives at the same time, so it's 5x faster.

There's some overhead compared to the raw file for the redundancy data, calculations and synchronization so it'll be a bit less, but you get the idea.

 

This has data for a single drive at 108MB/s writes, and a raidz1 of 5 of the same drive at 469MB/s.

 

https://calomel.org/zfs_raid_speed_capacity.html

 

and if I didn't care about data integrity, could it be just for speed? (each drive has bit of info) or is that too complicated or not really?

[LIGISTX]

2 hours ago, podkall said:

and if I didn't care about data integrity, could it be just for speed? (each drive has bit of info) or is that too complicated or not really?

RAID 0….

 

Me thinks you should just google RAID, and read for a few minutes  

[Blue4130]

2 hours ago, podkall said:

and if I didn't care about data integrity, could it be just for speed? (each drive has bit of info) or is that too complicated or not really?

That would be RAID 0. Data is written to all drives without the parity bit.  The downside is integrity. Loose one drive and all the data is gone.

[podkall]

2 minutes ago, LIGISTX said:

RAID 0….

 

Me thinks you should just google RAID, and read for a few minutes  

nah, RAID sounded pretty complicated to google

 

 

[LIGISTX]

2 minutes ago, podkall said:

nah, RAID sounded pretty complicated to google

 

 

I can’t tell if you are joking or not….

 

The Wikipedia page for RAID has had extremely easy to understand pictures for decades. It’s super simple to understand with a few pictures.