Mac NAS/Performance & Stuff

MacRS4

MacRS4

MacRS4

Soldato
Joined
15 Dec 2008
Posts
2,710
Location
London
A while ago there was some discussion on NAS (NAS's?) and around performance over 1Gbe/10Gbe etc. Anyway, something I'm working on includes some info that some of you nerds may find interesting. It's some performance testing/power consumption testing of a Synology DS923+ with 32GB RAM, 4 x Toshiba ENT 16TB Drives (SHR I think), 2 x Crucial P3 NVMe in a storage pool (you can't do this in the interface, you have to SSH in and do it) in RAID 0.

It's very nerdy, but includes some real test figures and some testing against encrypted folders, as well as tests over 1Gbe, 2.5Gbe and 10Gbe. Power estimates are based off 'my' usage model for my main work NAS that I use, however you can change the KwH costs and the layout of the usages.

Costs are retail however I didn't pay for any of this.

Anyway, the nerd sheet is here. I am aware it is dull aha. No real surprises in the performance really - 10Gbe is 10Gbe and decent on the spinners as well as the NVMe.

If you just want to see the perf figures, you can see those here.
Maybe some GiB/GB maths fudge going on at the minute, but it's broadly accurate. Pretty impressed with the unit - apart from it being awful for anything Plex related - no h/w video stuff (Quicksync?) I imagine being the problem there.
 
Thanks, good to see everything plays nice.

I'm keen to upgrade my setup back to Synology at some point this year (I had one, then went DIY and I'm regretting it). Would probably want an 8-bay with 10Gbe as default.
 
MacRS4 said:
If you just want to see the perf figures, you can see those here.
Click to expand...
I only have gigabit ethernet and spinning drives in my DS-1019+ with 8Gb RAM.

5Gb
Read - 109.5 MB/s
Write - 103.1 MB/s

1Gb
Read - 109.3 MB/s
Write - 101.5 MB/s

I did the tests with everything running as normal, shares were mounted, the NAS was chuntering away in the background, my torrent client is running so that could be doing stuff. My systems weren't sterile at all so these figures are very unscientific.

Interesting that in your case, your write speeds are slightly higher but in mine, the read speeds are a bit quicker but it's all pretty consistent.
My DS-1019+ has a pair of 1Gb Ethernet ports but my network is only gigabit so there's no point in me teaming the ports. My mechanical drives are 3 x 10Tb ST10000DM0004, 1 x 10Tb ST10000NM001G and 1 x 6Tb ST6000DM003.

I did measure the power usage a little while ago, it draws a little more than yours and that will be down to the extra drive. I have it automatically shut down just after midnight each night and then restart again at 08:30. I used to run Pi backups out to it overnight but changed all those to run on a Sunday.

I'm a lot happier with the Synology than I ever was with the Drobo.
 
Teaming would help with multiple devices. So 2 x 1Gbe to the NAS - you could get a 110Mb/s from one device and 110Mb/s from another to/from the NAS. There's also multi-channel SMB that can take advantage of it assuming you also LACP your host machine, but I've never really had much success with mSMB in reality.
 
Looks like a great bit of kit. One of the things I've been pointing out for this piece of work I'm doing is they want to give a team of remote workers a 4 bay DS923+ with 16 or 18TB drives - I think if they need that amount of storage they'd be better off with more bays and smaller drives tbh.

The rebuild time on 16/18TB drives is long at >1 day for a drive replacement running at full throttle where pretty much you can't use the NAS. At medium rebuild times where you can use the NAS and get degraded performance they're running at 4+ days for a rebuild. A build of a dataset for them (it's physics stuff for a University) takes about a week to do - the time exposure is huge for a drive failure.

I'm not sure that's been considered tbh.

I'm less concerned for 'my' stuff as mine are mirrored to another 4 bay NAS and I can use either as live, and they're back off elsewhere for glacier store.

Perhaps over-thinking this. A lot of people seem to treat RAID as a backup. I'd not want to spend well over a day in suspense at what may or may not happen to my data aha.

As a side note I don't know what the data is - I'd love to know, but they won't/can't tell me beyond profile stuff for performance testing. Given the uni it's for I imagine it's space stuff.
 
Also the write performance on RAID5 is a big hit, they should be using RAID10, or a proper enterprise array really. I'll have 4x14TB drives in RAID 5 which is mainly media and I can get back in a few days so I don't really care if it dies. The other 4x bays will be 1TB SSDs which is for Veeam/lab backups, photos and file shares, all of which is also backed up in B2.
 
Very happy with the DS1821+ so far, it saturates my 10 GbE connection to my SSD pool, can't properly test the main HDD pool as it's still expanding. I didn't quite realise how long it was going to take, I could have just set up a fresh R5 pool and re-downloaded my content quicker than it's taking to expand.
 
You must log in or register to reply here.
Back
Top Bottom