Thoughts on performance optimization

IF your questions pertains to using the file server performance, transfer issues, jumbo frames tuning, post it here.
Forum rules
This is a user forum for Synology users to share experience/help out each other: if you need direct assistance from the Synology technical support team, please use the following form:
https://myds.synology.com/support/suppo ... p?lang=enu

Thoughts on performance optimization

Postby Nuke288 » Tue May 15, 2012 8:20 am

Hi all,

I see a lot of threads about poor NAS performance, so I'd just like to add my $0.02 to the discussion - hopefully it will be of use to somebody.

When attempting to optimize the speed of your home network, the problem is more often than not with the infrastructure, not with the NAS itself (in my experience). A (data) chain is only as strong as its weakest link, so it can be useful to systematically analyze your network in order to locate the bottleneck.

The typical causes for poor performance are (in no particular order):
1) Switches and routers
2) Cables
3) Hard drives

Switches and routers
A transfer speed of around 11-12 MB/s is somewhat of a "magic number", as that is the top speed of a 100 mbit connection. This could indicate that you have a "fast ethernet" switch somewhere in your network, i.e. not a gigabit switch. Also, most consumer routers contain a 4-port switch, and many of these are limited to 100 mbit. It could also be useful to spend a little more $$$ for a high quality managed switch - I am personally very happy with my HP ProCurve 1810G switch.

Cables
Cable quality matters! If you have a poorly shielded ethernet cable somewhere, the two nodes connected by that cable will automatically downgrade themselves to 100BASETX, i.e. 100 mbit. This is from personal experience, as my Mac mini connected to a D-Link gigabit switch reported a 100 mbit connection, and it was all due to a dodgy ethernet cable. When I switched to a brand new 2m cat6 cable, the connection automatically jumped to gigabit speed. Try to use at least cat5e cables everywhere, or preferably cat6 cables. And don't buy the cheapest you can find! :)

Hard drives
Mechanical hard drives vary wildly in what kind of sustained speed they can deliver, and they do somewhat degrade with age. Even if your NAS is populated using brand new 7200 rpm drives (which it ideally should be), the limiting factor could still be that old and fragmented 5400 rpm drive sitting in your PC. Consider using a benchmarking tool to test each hard drive in isolation. Slow transfers due to hard drives are typically more evident when copying lots of small files, less so when tranferring one very large file. If max performance is key, having one or more SSD drives in your PC(s) will ofc be extremely useful.

Finally, I suspect Synology is "cheating" a bit when it comes to official max transfer speeds of their NAS products. Well, maybe not cheating - but at least these figures are achieved under optimal conditions. This probably includes using very fast SSD drives and link aggregation, and also a performance oriented RAID configuration - typically one that involves striping, like RAID 0 or RAID 10.

Example: My 1812+ is reported to have a max write speed of around 194 MB/s, but I've been unable to replicate this result myself - even using dual link aggregation and multiple transfers from fast hard drives (one of which is an SSD with a read speed of around 275 MB/s). The best I've been able to achieve is around 132 MB/s, and to get more I believe I would have to change my RAID type to something like RAID 10 (currently using SHR).
DS1812+ 4 x 3 TB Seagate Barracuda (SHR), 2 x 2 TB Western Digital (RAID 1), DSM 4.0
27" iMac 16 GB RAM, 3 TB Fusion Drive, NVidia GTX 680 2 GB, OS X 10.8 Mountain Lion
Mac Pro 6 GB RAM, 120 GB Intel SSD, 1.5 TB Western Digital, ATI Radeon 5770 1 GB, OS X 10.8 Mountain Lion Server
Mac mini 8 GB RAM, 2 x 500 GB HDD, OS X 10.8 Mountain Lion Server
User avatar
Nuke288
Rookie
Rookie
 
Posts: 31
Joined: Tue Apr 24, 2012 3:14 pm

Re: Thoughts on performance optimization

Postby henkan_2001 » Tue May 15, 2012 10:23 pm

Well written and very interresting. I have several thoughts of the actual performance using my brand new DS1812+.

Seems like my most limitting factor is my PC with a defargmented harddrive and a perhaps strange Windows 7 installation (since copying operation might slow down to 100 kb/s). When using a Dell laptop I was able to copy 90Mb/s from NAS to SSD.

What about the used RAID type? I have two volumes on my NAS.
Volume 1: SHB with 7 disks (3.5" 1.5TB) --> gives 70MB/s
Volume 2: Single 2.5" disk (2.5" 750GB) --> gives 90MB/s

The measured numbers are from a DS1812+ and a Dell laptop with internal SSD and Windows 7. I find it a bit strange that I get better performance when using the single 2.5" disk. Is the SHB volume really slower than a single 2.5" drive?

It would be very interresting to get some kind of list of internal rating between different RAID modes and questions like these...
* How should I configure my NAS for almost optimal performance?
* Would RAID-5 be faster than SHB on a DS1812+?
* Would a SHB volume with four disks be faster than SHB with eight disks?
henkan_2001
I'm New!
I'm New!
 
Posts: 7
Joined: Thu Jun 26, 2008 6:35 pm

Re: Thoughts on performance optimization

Postby TorontoTim » Wed May 16, 2012 4:50 pm

There's no guessing about Synology's test environment - they lay it out in 'fine print' and it's clearly designed for optimal performance (who can blame them!)

The client PC uses 2 x 60GB SSD's in RAID 0.

I suspect 99.9% of the people like me who get ~55MB/S transfer speed from the DS212 to the client PC over a GB network are writing to SATA HDD's that are likely a couple years old etc. etc., not state of the art SSD's chosen and configured specifically for writing speed.

From their Performance tab:

http://www.synology.com/products/perfor ... =us#tabs-2

Test Environment:

Server:

Operating System: DSM 4.0
HDD: WD2003FYYS 2TB HDDs (DS411slim uses WD5000BEKT 500GB HDDs)
Volume Type: Basic (1-bay models), RAID 1 (2-bay models), RAID 5 (all other models)
Network Environment: 1Gbps LAN; MTU 1500; directly connected to the client PC (DS1512+/DS1812+/RS2212(RP)+/DS2411+ with 3GB RAM, DS412+ with 1GB RAM, connects to HP switch 2824 for Link Aggregation testing)

Client PC:

For All Testings: Intel Core i5 750 2.67GHz; 4GB DDRIII; SVP200S3 (60GB) SSD x 2, RAID 0 ; Intel Gigabit CT; MTU 1500; Windows 7; Transfering single 10GB file (For Link Aggregation); Ubuntu 9.10; Apache v.2.2.12 (Web Server Responsiveness Testing)
TorontoTim
I'm New!
I'm New!
 
Posts: 2
Joined: Sat May 12, 2012 6:45 pm

Re: Thoughts on performance optimization

Postby Nuke288 » Fri May 18, 2012 10:31 am

Good point about Synology publishing details about their test environment; I didn't bother to look for that information, I just assumed it was optimized for performance - which it clearly is. Dual SSDs in RAID 0 is about as fast as you can get, so they clearly want no bottlenecks on the client side. Ofc it makes perfect sense for them determine the max transfer speed of each NAS under ideal conditions.

My experience has been that my transfer speed hovers around 60-70 MB/s when using my (pretty old and fragmented) 1.5 TB internal SATA drive, but it jumps to 120+ MB/s when I use my SSD. This SSD is only 120 GB, and it's used mainly for the OS and all my applications, so there's really very little room for Blu-ray rips etc. Time so save up and buy an additional SSD, I guess :)

I also find it interesting that Synology use a default MTU of 1500 in their test scenarios. This could indicate that it is pointless to fiddle around with Jumbo Frames.

Regarding RAID types, I note that Synology use RAID 5 to test multi-volume (i.e. >2 bays) NAS products. Even though I have read in other threads that there is no significant performance penalty associated with Synology Hybrid Raid (SHR), it might be interesting to test performance under RAID 5. Then again I enjoy the possibility of mixing hard drive capacities within the same volume, and 120 ish MB/s is more than ample for my home network needs.
DS1812+ 4 x 3 TB Seagate Barracuda (SHR), 2 x 2 TB Western Digital (RAID 1), DSM 4.0
27" iMac 16 GB RAM, 3 TB Fusion Drive, NVidia GTX 680 2 GB, OS X 10.8 Mountain Lion
Mac Pro 6 GB RAM, 120 GB Intel SSD, 1.5 TB Western Digital, ATI Radeon 5770 1 GB, OS X 10.8 Mountain Lion Server
Mac mini 8 GB RAM, 2 x 500 GB HDD, OS X 10.8 Mountain Lion Server
User avatar
Nuke288
Rookie
Rookie
 
Posts: 31
Joined: Tue Apr 24, 2012 3:14 pm


Return to File Server Performance

Who is online

Users browsing this forum: kisraelsen and 2 guests