DSM-G600, DNS-3xx and NSA-220 Hack Forum

Do you also know the power consumption with the drive spinning?
And does it use more power during reading/writing?

I had planned to test power consumption with my Conceptronic CH3SNAS for a while now, unfortunately I never came around buying a Kill-A-Watt. I currently only have on 500GB Seagate drive in there.
Still, very interesting  to know. Almost no review mentions power consumption for always-on devices.

Today I tested the power consumption of the above setup (two Seagate 500GB 7200 RPM drives).

Unit turned off:                       6 watts
Unit turned on, drives spinning idle: 26 watts
Unit turned on, drive heads seeking:  29 watts
Unit turned on, drives not spinning:  14 watts

The results were measured with two different consumer-grade (cheap) watt meters at the wall, and the numbers where averaged. I live in a 230 volt country.

Note: The results may be a watt or two off compared to the standard DNS-323 because I installed a bigger fan.

If you want to know how much it costs you having the DNS-323 turned on 24/7 (or for shorter periods) with or without using drive spin down, you can use this Javascript power cost calculator.

Last edited by oken (2008-03-15 22:24:37)

Can anyone else confirm a speed increase with the firmware?

I  have not updated yet since my system is working flawlessly right now and I don't have a compelling reason to upgrade... unless of course there is a bump in speed

I have two of the 750 GB Green Power drives in a RAID 0 config and I am seeing the same performance as blahsome on a GB LAN more or less with FW 1.04

I use RAID 0 because I hate to lose an entire drive to RAID 1... besides I back up my home PC's (Linux and Vista) to the NAS so if a drive fails in either the PC's or the NAS there is a copy somewhere.  I know if my house burns down I am out of luck but if that happened I would have more pressing problems

I currently own a DLink switch DGS-2208 that supports jumbo frames.
I also have two GB capable PC's a laptop (part time) attached to the switch... everything else Xbox360, DirecTV receiver, and printer are all 10/100
So no jumbo frames for me since the 10/100 clients would have problems

Last edited by ithrowpicks (2008-05-18 04:34:22)

Everthing that is on the NAS is a copy of what's on the PC's except for a few of movies I ripped and placed on the NAS so my Popcorn Hour unit can access them without leaving a PC on all the time adding to my electric bill.

If a drive in the NAS died, after replacing the drive I would run a full backup from the Vista PC and from the Linux box create a zipped tar file of the home directories and I am back in business.   That's why I mentioned if both my PC's and the NAS were destroyed at the same time I would lose a lot but since the chances of losing everything is pretty slim  I get to have 1.5 TB (RAID 0) of storage on the NAS instead of 750 GB (RAID 1).    Women like it when you say I have a terabyte of NAS storage space

Besides, I don't know about you guys but I don't have a spare 750 GB drive laying around anyway so if I used RAID 1 I would still have to wait until I find a good deal for another drive of same type and have it shipped.  So while that is happening I am running the risk on losing data on the one remaining drive is something goes wrong.

The videos I would have to rip again but that's no big deal since I only rip them and put them in iso's.  I don't convert them to .mkv or .avi's which would obviously take more time.

Thanks for the words of caution... they have not fallen on deaf ears

But back to the original question... has anyone else noticed a speed increase with 1.05?

I compiled my test scripts into a small GUI application, if anyone wants an easier way to benchmark and compare results:



It can be downloaded here NAS Performance Tester 0.1

The source code (in C# .Net 2.0) and a batch file for compiling it is included in the zip file, if you are uncomfortable (you really should be) with running unknown programs downloaded from a random guy's website. The code is hereby released to the public under the CC Attribution license. In other words: Feel free to hack/modify and re-publish it, just keep a reference to the original version in it.

Note 1: the result in the screenshot is just show-off. I can only hit 27MB/sec read when i connect the DNS-323 directly to my workstation with a crossover cable and running jumbo 9000 on both. If I run through my switch with jumbo frames disabled, I get something in the ballpark of 18MB/sec read and 16MB/sec write.

Note 2: The program uses the SI definition of Megabytes (1 MB = 1.000.000 Bytes).

Thank you, fordem.

By the way, don't worry about finding some files of the right size. The program will automatically generate some temporary files of the specified size (200MB is a good starting point) and use them for the timed transfers. Note that the the read benchmark will take a bit longer than the write benchmark to complete. That is because of the way the test files are first generated and written to the NAS (the method prevents the files from being cached before being read back. Unfortunately, the method is also a bit slow).

So let us see some comparable benchmark results from more people. The result output of the program can be copy/pasted as text, so you don't have to take screenshots.

oken you rule. awesome app!

here are my test results, dns-323, firmware 1.05, raid0, nvidia nforce2 gigabit ethernet onboard, jumbo 9k

Running 5 iterations of a 500MB file write benchmark on drive M:...
(Jumbo 9k, with bridge)
Iteration 1:     13,26 MB/sec
Iteration 2:     13,33 MB/sec
Iteration 3:     13,32 MB/sec
Iteration 4:     13,34 MB/sec
Iteration 5:     13,34 MB/sec
----------------------------------------------
Average (W):     13,32 MB/sec
----------------------------------------------

Running 5 iterations of a 500MB file read benchmark on drive M:...
(Jumbo 9k, with bridge)
Iteration 1:     18,13 MB/sec
Iteration 2:     18,04 MB/sec
Iteration 3:     17,86 MB/sec
Iteration 4:     18,19 MB/sec
Iteration 5:     18,15 MB/sec
----------------------------------------------
Average (R):     18,07 MB/sec
----------------------------------------------

i was like wtf but then i remembered i have a freaking bridge (lan-wlan) which i don't even use anymore (wlan is disabled). deleted the bridge, ran the tests again and voila!:

Running 2 iterations of a 500MB file write benchmark on drive M:...
(Jumbo 9k)
Iteration 1:     19,21 MB/sec
Iteration 2:     19,44 MB/sec
----------------------------------------------
Average (W):     19,32 MB/sec
----------------------------------------------

Running 2 iterations of a 500MB file read benchmark on drive M:...
(Jumbo 9k)
Iteration 1:     28,52 MB/sec
Iteration 2:     28,67 MB/sec
----------------------------------------------
Average (R):     28,6 MB/sec
----------------------------------------------

i was amazed what this little box can do. and the best thing is i'm not even sure my switch supports jumbo (level one gsw-0506)!
anyhow, i'm buying another dns-323 next week for music this time...

Thank you richneo and SE

SE: Very nice, you beat my 27MB/sec read result, but then again you are running raid 0 I wonder if a single/stand-alone or JBOD drive setup would have the potential to be even faster. At these speeds, the bottleneck should still be the unit, not the drives themselves.

By the way, I updated the program about an hour ago after having tested it on my Vista box. It didn't work properly if you didn't run it with administrator privileges, but that have been fixed now (by changing the location the temporary test files are written). So if you tried it on Vista and it didn't work for you, just re-download it and it should work.

first 2 test is for dns-323 with standard drive settings and jumbo frame 7000 and the last is dns-313

Running 1 iterations of a 200MB file write benchmark on drive y:...
Iteration 1:     19.22 MB/sec
----------------------------------------------
Average (W):     19.22 MB/sec
----------------------------------------------
Running 1 iterations of a 200MB file read benchmark on drive y:...
Iteration 1:     25.75 MB/sec
----------------------------------------------
Average (R):     25.75 MB/sec
----------------------------------------------
Running 1 iterations of a 200MB file write benchmark on drive x:...
Iteration 1:     18.85 MB/sec
----------------------------------------------
Average (W):     18.85 MB/sec
----------------------------------------------
Running 1 iterations of a 200MB file read benchmark on drive x:...
Iteration 1:     26.95 MB/sec
----------------------------------------------
Average (R):     26.95 MB/sec
----------------------------------------------
Running 1 iterations of a 200MB file write benchmark on drive z:...
Iteration 1:     8.31 MB/sec
----------------------------------------------
Average (W):     8.31 MB/sec
----------------------------------------------
Running 1 iterations of a 200MB file read benchmark on drive z:...
Iteration 1:     18.18 MB/sec
----------------------------------------------
Average (R):     18.18 MB/sec
----------------------------------------------

I don't think you're missing any settings - but more likely just overlooking one thing - the other end of the transfer.

A file transfer requires that you move a file from one disk to another, across a network - sounds simple, but it's really a little more complicated than that ...

It's more like - read from disk_1, through disk_interface_1, across expansion_bus_1, through network_interface_1, across the network, through network_interface_2, across expansion_bus_2, through disk_interface_2, and finally write to disk_2.

In these comparisons, one end is more or less static - the DNS-323 - but the other end can be dramatically different.  I ran oken's utility from three different systems, all equipped with integrated gigabit ports and had different results from all three.

I should have notes that I can post - probably tomorrow - but going from memory, my desktop was significantly slower than either my laptop or my server, and the perhaps most noticeable difference in hardware was the system memory - the desktop is a 3GHz P4 with 512MB RAM and a Broadcom nic, the server is also a 3GHz P4, but this time hyperthreading, 1GB RAM and an Intel nic and the laptop is a 1.6 Pentium M, 2GB RAM and a Broadcom nic.
=============================================================================================
Edit - I'll just post the promised results here

System #1

IBM xSeries 206 server - 3 GHz P4(HT), 1GB RAM, Server 2K3 R2 SP2, 2x250GB SATA disks in RAID1, IntelPRO 1000 gigabit card, no jumbo frame

Average write (200MB) - 15.65 MB/sec, average read - 18.78 MB/sec, Average write (2GB) - 12.97 MB/sec, average read - 20.5 MB/sec

System #2

Dell Optiplex GX280 - 3 GHz P4, 512 MB RAM, XP Pro SP3, 80GB SATA drive, Broadcom NetExtreme gigabit card, no jumbo frame

Average write throughput - 6.75 MB/sec, average read throughput - 14.92 MB/sec

System #3

Dell Inspiron 600m - 1.6 GHz Pentium M, 2GB RAM, XP Pro SP2, 80GB EIDE drive, Broadcom NeteXtreme gigabit card, no jumbo frame

Average write throughput - 14.66 MB/sec, average read throughput - 14.53 MB/sec
==============================================================================================

As you can see - the server gave the best results, the laptop was slightly slower and the desktop was dismal.

The point here is that you must know what the other end of the transfer is - and that even things like disk fragmentation will impact your results.

Oh - with jumbo frame (9000) the server will give 17.64 MB/sec average write and 28.68 MB/sec average read - with a single best result of 30.33 MB/sec, smaller frame sizes gave varying results, write speeds went up, read speeds went down - the best write speed achieved was 18.92 MB/sec average (19.13 MB/sec single best) with frame size at 4000.

Last edited by fordem (2008-05-22 15:11:00)