Post your performance results

Hello All,

Just got the new RNDP6000, setup real quick with 4-1TB Seagates and dual NIC in Adaptive Load Balancing mode: I always use Intel's NAS Performance toolkit over IOMETER. It is also connected to a Netgear GS116 Gigabit switch. I like the NAS Perf Tool as it shows consistent results and performs a variety of tests, of course with any test software, the better the host computer, the results will be slighty better.

The test was performed from a i7-930 workstation with 12GB RAM, single NIC. Next will use a teamed NIC to compare results.

Anyhow here are the results: MB/s

Test: HDVideo_1Play Throughput: 59.110
Test: HDVideo_2Play Throughput: 55.416
Test: HDVideo_4Play Throughput: 53.983
Test: HDVideo_1Record Throughput: 113.548
Test: HDVideo_1Play_1Record Throughput: 72.324
Test: ContentCreation Throughput: 1.038
Test: OfficeProductivity Throughput: 35.621
Test: FileCopyToNAS Throughput: 104.949
Test: FileCopyFromNAS Throughput: 46.685
Test: DirectoryCopyToNAS Throughput: 3.207
Test: DirectoryCopyFromNAS Throughput: 4.964
Test: PhotoAlbum Throughput: 10.666

I have a ReadyNAS Pro with 6x Constellation ES.2 Seagate 500G drives. It's running standard 1500B frames and has XOR NIC teaming enabled, to a cheap gig switch. I have a new Mac Pro connected to the same switch with a single gig interface. I have maxed out the RAM.

File copy of very large files from Mac Pro to the ReadyNAS consistently exceeds 825Mbps over AFP. This is not the optimized "Mac-only" filesystem setup. This is presumably a NIC and cache test, I'm guessing write performance is lower.

G.

My results from a Core i3 450 PC with 8GB RAM via gigabit connection on a Netgear GS108 switch to my ReadyNAS Ultra 6 running RAIDiator 4.2.15.

3 x WDC WD2002FYPS 2TB drives in X-RAID2.

Settings:

Enable disk write cache: Enabled
Disable full data journaling: Enabled
Enable jumbo frames: Disabled
Enable fast USB disk writes: Disabled

Network shows 0 errors.

IOmeter Read MBps = 5.255977
IOmeter Write MBps = 67.971737

Write MBps = 40 MB
Read MBps = 32 MB

Jetlag wrote:

I wasn't aware that this was a competition. (?) These results are slower than what I was getting at my previous residence where I was able to run lots of dedicated Belden Cat6 lines everywhere before the sheet rock went up. Those cables were run optimally; no kinks, tension or tight corners and they were kept away from line voltage wiring and other potential interference. All terminations were also Cat6. In my new residence my PC now shares a single ~60 foot Home Depot quality Cat5e with my laptop, a printer, an IP Camera and a Slimdevices music streamer. These are all plugged into an inexpensive D-Link switch. An electrician without any apparent network training installed the Cat5e wire, and based upon his horrible terminations (all of which I had to re-terminate) I assume these cables were not run with much attention to detail. Here is an example of his work: One other issue with this single cable is that it and two other network cables accidentally got cut while work was being done in the utility room. Since it was impossible to replace the cable runs in the finished house I fixed them using Cat5e junction boxes. Not an optimal solution by any means. This single cable then connects to my HP switch which is also connected to my Pro. So yes, I have some nice gear but I am obviously running the speed test over marginal cabling. I also cannot enable jumbo frames. For some reason my old Linksys AP stops working if I enable JF on the rest of the network. I plan to replace it soon with a new HP V-M200 AP, just waiting for some reviews to get posted first. I even tried flashing my old AP with Tomato and DD-WRT with the same results. If you have any suggestions as to how I can improve my results on this setup I am open to suggestions. Yes I would love to run new cables, simply not an option unfortunately.

I've seen every type of cable nightmare. I love it when old school telephone guys terminate high-speed Ethernet - what a mess.

Since you're apparently handy with a 110 punchdown tool I'd pull those wires off that wall connector and retwist them nice and tight from the cable right up to the connection points. The only twists that really matter for bandwidth purposes are color for color - blue twisted with blue/white, et cetera.

H.

ReadyNAS Ultra 6 Plus
RAIDiator 4.2.15
X-RAID2, 2 disks, 32% of 1751 GB used
Seagate ST32000644NS 2TB x 2

NetGear GS724T-300NAS Switch (wtf is the NAS?)

(Old) Home Built PC using Intel Dual Port Gigabit NIC PCIe card.
Intel Core2 Quad Q9300 2.5GHz
Raid Mirror 2 x Seagate Barracuda 7200.11 1TB drives

LAN Speed Test (02-03-2011 at 22:02:29)
2GB file

                        ---Writing---     ---Reading--- 
Packet length   :      2,097,152,000     2,097,152,000
Time to complete:      55.9666840        34.6429079
Bytes per second:      37,471,436        60,536,258
Bits per second :      299,771,488       484,290,064
                       -------------     -------------
Mbps:                  285.8843689       461.8549957

Teaming / LAG / Trunking / whatever enabled over 2 ports PC = Switch = NAS.
Jumbo frames NOT enabled. What a bloody nightmare. Maybe with phone tech support I can reconsider.

I also ran Iometer tests: Reads 30-50 MBps, writes 72-83 MBps.

Drag and drop 3.12 GB of Metallica in a folder NAS -> PC 49,328,385 Bytes per second
Drag and drop 3.12 GB of Metallica in a folder PC -> NAS 79,865,004 Bytes per second

The drag tests are self timed, no cache cleared so rather unscientific.

Azethoth what brand and model disks are you using? Are they on the HCL? How many disks do you have installed?

What version of RAIDiator x86 is installed on the ReadyNAS?

Are you using X-RAID2? If so, the default single-redundancy or the optional dual-redundancy?

The -300 indicates it's the 3rd hardware revision. The NAS indicates that the product is for the North American region and the language of the included documentation.

What disks do you have installed in your PC? Are you using RAID in your PC?


The kind of NIC bonding (http://www.readynas.com/bonding_en-us) used is important too. Do you get better or worse performance if it's disabled?

Updated the post a bit. Just single redundancy so far on the X-RAID2. I plan on increasing the number of drives later.

Haha yeah the drives are now strictly HCL. I was dumb and got some WD green drives when I first ordered it.

I am using IEEE 802.3ad LACP. I believe I tried to get it to double throughput for all 3 participants.
The Intel dual port card is using Maximum Bandwidth for the LAG setting. I did not mess around with any processor offloading yet.
The switch is using LACP as well. The Ultra 6 Plus is set to LACP as well.

Performance seems improved with LAG. I will need to rerun the non LAG tests again to be sure as my network just changed dramatically with the switch added in and Cat 7 cables (7') replacing a melange of cat 5 - cat5e in a variety of lengths and quality.

I can also do some direct connect runs maybe if that would be of interest, although I would never use it in such a configuration.


Mmm, interesting, I ran the Lan Speed test at 8GB thinking it would be slower due to less caching effects but:

                        ---Writing---     ---Reading--- 
Packet length   :      8,388,608,000     8,388,608,000 
Time to complete:      191.8078302       121.3541888 
Bytes per second:      43,734,440        69,124,998 
Bits per second :      349,875,520       552,999,984 
                       -------------     ------------- 
Mbps:                  333.6672974       527.3818817

There are multiple factors at play here. Take a look at ReadyNAS Performance Expectations.

There isn't a resync still under way is there? While the drives are syncing sector-by-sector performance is slower.

With a single 7200rpm disk you might expect to achieve 60 MB/s.

Using RAID-1 (or X-RAID2 with 2 disks) changes things up a little. However you have to remember that writes to either machine have to be written to both disks. If you take a look at the Performance section of the Definitive Guide to the ReadyNAS Pro (a machine with the same CPU and with comparable performance) you'll see that NetGear's setup for optimal performance involved having all drive bays in the ReadyNAS full. Presumably they used a multiple disk array in the PC as well.

LACP is a good teaming mode to use. However with your current configuration of disks you can't take advantage of it. You should find that with a RAID-5 array (3+ disk array with single disk redundancy) in each machine that performance is better.

There is a good community addon for testing the internal disk performance in your ReadyNAS called Bonnie++. Obviously with network overheads performance is going to be slower over your network, but it does give you an idea of what the disk configuration you have is capable of.

Dell R710 Server, ReadyNAS 3100, GS748Tv3
Jumbo Frames, LACP, MPIO (iSCSI), Flow Control disabled

CIFS: 107 MB/s read, 112 MB/s write
iSCSI: 117 MB/s read, 105 MB/s write