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

Sparkss - if you see, as you are, a performance hit when moving from 100 mbps to gigabit - the cables are, more often than not, the problem.

blahsome - that seems to be one of the controversies of jumbo frame.

The way I understand it is that for jumbo frame to work the entire network path between the two endpoints must support the frame size you intend to use - it seems to me that some people misenterpret "entire network path" to mean the entire network, whilst I see it as only that portion of the network between the specific endpoints

I don't have that much personal experience with jumbo frame, but, the little that I have says jumbo frame can co-exist with standard sized frames - there is some negotiation that establishes MSS, or maximum segment size - as an example when you're configuring the DNS-323 for jumbo frame, you can configure either end first, it will continue to communicate even though the two end points have different settings.

I did some tests with oken's benchmark utility and got the following results - please note the switch I was using at the time does not support jumbo frame.

standard frame on both ends - average write - 16.43, average read - 20.13
9000 byte frames on the DNS-323, standard frames on the other - average write - 12.24, average read - 20.02

Update -

In the months after this was posted I acquired a switch that supports jumbo frame and for several months have been running a network with a mixture of 100mbit and gigabit clients both with and without jumbo frame enabled - those that support jumbo frame have it enabled, those that don't, dont - along with 802.11g wireless.

The system I sit at all day long - most of the day anyway - is gigabit without jumbo frame and the databases I work with are on an IBM server that is gigabit with jumbo frame - in day to day usage I have noticed no adverse effects - I can also access the same databases using a wireless equipped laptop, it is noticeably slower, but I would think that is to be expected when you move from a gigabit network to an 802.11g network.

I'd say - it IS possible to mix jumbo frame and non jumbo frame on the same network, you can expect some degradation (see the figures from my earlier tests), but, it is, in my opinion quite usable.

Last edited by fordem (2008-11-25 00:26:27)

Great. Thanks, guys.

It seems that some experimentation is in order. Of course, I'm still waiting for the RMAed DNS-323 to come back.

Thanks oken - I did find the lack of a scroll bar a minor irritant yesterday.

Now, if I may be so bold as to suggest another change - at start up perhaps you could have it check for the presence of the test files and delete them if they are found, I know you delete them at the end of the test, but if the test fails for any reason (network failure perhaps ) the test file remains and the next test run will fail with what is essentially a file already exists error.

A cancel button would also be nice, even if it only cancelled the remaining iterations, as would a way to add a comment to the "logged" results.

In addition to the need to be aware that the test computer can skew the results - if the DNS-323 is in power save mode, the first iteration of the write test will be affected by the spinup delay, and this of course will drag the write average down.

for the cancel button thingie, you should add Application.DoEvents at the beginning of every for loop of iteration, that way, the application will seem unresponsive while doing the test but it will process messages in the beginning of next iteration..

another solution would be threading

Thank you for your feedback on the programming issues, SilentException and sjmac.

I have implemented a simple cancel button (that completes the current read/write and stops afterwards) using a conditional break statement in the main for loop with an Application.DoEvents() in the beginning of the loop.



NAS Performance Tester 0.4

The cancel button works kind of OK. You have to bang it a few times for the event to register because of the program's unresponsiveness while the temp test file generation loop is running. That could probably be fixed with threading, but that is currently truly out of my league.

Last edited by oken (2008-05-24 12:33:53)

they are removed by the app

Hello,

I'm a noob in networking but I hope you can help.

When I am writing files to the disk I noticed that transfer speeds were very low, that's why I decided to try the nas tester. After getting over the 94% error I got my disk to work but...

The transfer speeds on my system are very low...

Write I got a 0.73MB...

Then shortly after the app (nas tester) crashes.

I am running the DNS 323 v 1.7 on 1.5 TB seagate
The DNS is conected to us robotics sureconnect

http://www.usr.com/images/products/9105/9105.jpg
linked to wireless
http://images.hardwareinfo.net/products … 052149.jpg
and in my P4 3.2  4 Gig RAM
http://lib.store.yahoo.net/lib/directron/USR541705.jpg

Is it possible that it are due to the components used or is it a setting problem? What do I have to do in order to get better results.

Thanks,

Greets,

Nicolas

(Edited by fonz: Please do not embed external images with unclear or non-free license)

Thanks for your reply , I suspected something like that and did the test today:

Running warmup...
Running a 200MB file write on drive Y: 5 times...
Iteration 1:     8,09 MB/sec
Iteration 2:     8,11 MB/sec
Iteration 3:     8,13 MB/sec
Iteration 4:     8,13 MB/sec
Iteration 5:     8,14 MB/sec
------------------------------
Average (W):     8,12 MB/sec
------------------------------
Running a 200MB file read on drive Y: 5 times...
Iteration 1:     8,44 MB/sec
Iteration 2:     9,07 MB/sec
Iteration 3:     9,01 MB/sec
Iteration 4:     9,22 MB/sec
Iteration 5:     9,05 MB/sec
------------------------------
Average (R):     8,96 MB/sec
------------------------------


I suppose these test results are normal seen the configuration.

Are there wireless routers who can get to 1gbit?

I think we can conclude that on a gigabit network the 323 is good for between 15 and 17 MB/sec, best case.

Does anyone know how the less expensive QNAP NAS boxes compare?  The recently announced TS-410 is getting into a similar price range of the four disk version of the 323 and it *appears* to have more features and a more sophisticated management environment.  Anyone know how well the QNAP boxes do in the throughput department?

I apologize for re-animating this old thread, but I just wanted to announce the release of a newer version of the NASTester utility. It has now been rewritten with proper threading, so the GUI is more responsive, among other things. The core copy and timing algorithms remain the same, so the benchmark results from version 0.4 and 1.0 are still comparable.

Like before, the new version can be downloaded here: http://www.808.dk/?code-csharp-nas-performance

Thanks Oken - I missed the 1.0 release somehow, and I've been using the 0.4 release occassionally and the "lack of a screen update whilst running", which I believe will be fixed with the addition of multithreading has been a source of minor frustration.

The new larger file size is also most welcome - keep up the good work.