In reply to a post by Woolwich:

The NAS gives me 74.06M bytes/sec for the 1GB and 71.59M bytes/sec for the 512 MB

The RAID 5 was 42.11M bytes/sec for the 1GB and 42.95M bytes/sec for the 512 MB.

I was thinking of making more use of the RAID as I expected it to be quicker therefore better when working with large files on my Mac. Maybe I'll just keep it as a backup. In this case RAID is a backup!

It's well known that DAS (Direct Attached Storage) is usually faster than NAS/SAN traffic. eSATA/USB3 can do 200MB/s + (pure USB3 max theoretical is 625MB/s)

Its not the speed of the traffic over the network slowing the NAS data rate, but the speed in which it writes to the drives. Stick a 500mb/s SSD in the NAS and it will probably be faster - but spinning disks aren't that great from a transfer speed point of view.

The location is not really relevant - 1m or 90m will not show any difference.

The NAS is potentially limited by the drive speed/capability. 74.06 and 71.59 and much the same and if you repeat teh tests they could switch round or get closer.

The RAID5 could well be down to the drive firmware having a processing limit and pushing te data to each drive as well as calculating parity and inserting that. And parity could be another 35% meaning the drives are storing at around 56 MBps.

In reply to a post by haydnwalker:

It's well known that DAS (Direct Attached Storage) is usually faster than NAS/SAN traffic. eSATA/USB3 can do 200MB/s + (pure USB3 max theoretical is 625MB/s)

Its not the speed of the traffic over the network slowing the NAS data rate, but the speed in which it writes to the drives. Stick a 500mb/s SSD in the NAS and it will probably be faster - but spinning disks aren't that great from a transfer speed point of view.

So I was right to be surprised, the RAID should have been faster. The NAS is a Synology running SHR with 7200 drives. The RAID also has 7200 HGST drives.

Simple mirroring of pairs of drives RAID 1 or simple stripe of mirrors (RAID 10)
both provider faster and easier restoration than any parity-based schemes.

With increasing capacity per £ may become favoured in the long term over the risk and complexity burden of other methods.
With any parity-based scheme the time taken to rebuild a disk and parity can exceed the time before the next disk failure; it's not just theoretical and it does happen that people lose a whole set that way with a disks of similar age living in same environment for years.

There is also a distinct advantage in not being dependent on quirks of any particular RAID drive controller and being able recover some usable data from any single drive of a mirror.

I like the idea of Synology devices but the affordable ones don't by any means pack a powerful CPU which needs to be factored into choosing for scenarios suited to capabilities.
Marketing of NASes would like to offer you about an ever increasing range of add-ons or apps while avoiding that level of detail though, and that's true of QNAP and most others unless you build something yourself.

Edited by prlzx (Mon 12-Jul-21 21:16:28)

In reply to a post by prlzx:

Simple mirroring of pairs of drives RAID 1 or simple stripe of mirrors (RAID 10)
both provider faster and easier restoration than any parity-based schemes.

I don't disagree with what you say (other than the price of hard drives!).

The options are proprietary NAS or roll yer own from 'free' software. I gave that a go with a Centos based server distro and although it had a GUI it was a pain to learn, use when command line was needed and get any sensible instructions/support. I never did work out how to mount external hard drives so the idea or adding RAID was a non starter. All fine if you do it daily, your average punter who wants a big centralised hard drive just wants it to work.

In reply to a post by Woolwich:

I did some tests.
...snip...
The NAS gives me 74.06M bytes/sec for the 1GB and 71.59M bytes/sec for the 512 MB

The RAID 5 was 42.11M bytes/sec for the 1GB and 42.95M bytes/sec for the 512 MB.

I did some more tests, this time copying the files from the NAS and RAID to my Mac. So kinda testing read speeds I guess.

The results surprised me - again!

From the NAS to the Mac, 82.61M bytes/sec and 79.55M bytes/sec for the 512 MB and 1GB files respectively.

From the RAID to the Mac, 119.32M bytes/sec and 126.34M bytes/sec.

So the RAID is much better reading than writing. And that's RAID 5, I was half thinking of reformatting it as RAID 10 (or 1+0?). That would be a bit quicker yet?

So far that's Big Files, cos that's what you lot say to test with. In Real Life I read and write to hundreds of 4 or 5K files (kilobytes, not 4K TV!). Anyway, it seems if I want to read fast I should use the RAID, write fast, use the NAS...

The NAS figures are not much different to before - so almost certainly teh drive is limiting.

The RAID - it has to be te controller or firmware that slowed down the write and the read speeds are maxing out the Gigabit Ethernet. 125MB * 8 = 1000Mb or 1 Gb

This is a good overview:

https://blog.storagecraft.com/raid-performance/

Distilled into one salient remark, “In terms of RAID, reading is extremely easy and writing is rather complex. Read performance is effectively stable across all types. Writing, however, is not.”

Much like my comment on Monday.

In reply to a post by MHC:

The RAID - it has to be te controller or firmware that slowed down the write and the read speeds are maxing out the Gigabit Ethernet. 125MB * 8 = 1000Mb or 1 Gb

Yes but no. The RAID is the one connected via USB3.1 to an eSATA port (using convertor cable).

As said above

In reply to a post by haydnwalker:

It's well known that DAS (Direct Attached Storage) is usually faster than NAS/SAN traffic. eSATA/USB3 can do 200MB/s + (pure USB3 max theoretical is 625MB/s)

So not bad really.