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

Dunno - I have a DNS-343 w/1.03 firmware and I found my Realtek (WinXP) driver completely bombs w/Jumbo Frames while another machine with an Intel Pro/1000 MT NIC cruises along around 310Mb/s....

So don't discount all the hardware in between.

-Ben

The title of the thread is file transfer speed is slow, yet the entire discussion is on network speeds - I just want you to recognise that the two are not the same - the file transfer speed will be limited by the network speed, but, the network speed is not always the limitation in a file transfer.

Compare the network to a water pipe, you may have a 48" water main that can flow 15,000 gals/hour, but if you connect it to a pump capable of pushing 300 gals/hour, you won't get the 15,000 gals you're expecting.

jocala wrote:

It seems that there's no pattern, just that some folks gigabit speed sucks and others don't. Mine sucks.

Perhaps the lack of a pattern is caused by not looking in the right place for the pattern.

It stands to reason that if some folks can get decent gigabit speed that the unit is capable of delivering decent gigabit speed, and if other folks can't match that performance, then perhaps they need to look at what is different, and in posts like yours - not even discussed - the disks from which & to which the files are being transferred

Here's a nice article to put that in it's proper perspective - http://www.tomshardware.com/reviews/gig … ,2321.html

Oh - for what it's worth - I've found the performance of the Realtek based gigabit NICs to be subpar, and more so if used with Microsoft supplied drivers - gigabit cards delivering speeds in the 70 mbps region - speeds that I would consider embarrasingly low , even of a 100mbps NIC - simply switching to the card manufacturers' supplied drivers or Realtek's drivers (Realtek manufacture chipsets which are used by other companies to build NICs or to integrate in their system boards) will often double the throughput, but, 2 x 70mbps is only 140mbps, still ridiculously low.

If you'd like to be able to measure network throughput, rather than the speed of the entire chain, google for QCheck - it's a free utility that runs entirely in memory and can be used to measure a transfer without it being affected by the speeds of the disks or the disk interface.  By the way - they also offer a range of endpoints compiled for different operating systems and processors, including one that runs on the DNS-323 and should run on the DNS-321 - the DNS-323's network interface is capable of transfers exceeding 400 mbps, I shoud think the DNS-321 wouldn't be that much different.

I could be wrong but there's probably a way to test - QCheck is just a way to manipulate the endpoints, and there are endpoints for linux and quite likely which ever Mac OS you're running.

Now - about that comment ...

jocala wrote:

I would have thought the straight connect between a Mac Mini and the DNS-321 would have eliminated network bottlenecks.  Surely Apple's hardware wouldn't be so terrible as to bring the 321 to it's knees

I can't help but wonder if you took the time to read the article I linked to - if you did, you missed the point, and even missed the point of my water pipe analogy.

You've eliminated the network as the bottleneck - so what does that leave you with?  The Mac & the DNS-321.

As per my previous post ...

It stands to reason that if some folks can get decent gigabit speed that the unit is capable of delivering decent gigabit speed, and if other folks can't match that performance, then perhaps they need to look at what is different, and in posts like yours - not even discussed - the disks from which & to which the files are being transferred

The problem just might be in your Mac

Before you dismiss the idea, try researching it - you just might be surprised.

Maybe you should gotten a DNS-323 - I get 24~30MB/s from mine

Sooner or later, though, you'll come to the realisation that maximum throughput is vastly different from real world usage.

Smallstack,
Maybe you can help with a clarification please (even though it seems clear from your post)? 

My DNS-323 units are set to jumbo 9k.  Max setting on my 1,000mb network cards is jumbo 7k.  Portables from time to time are connected in, and they support max jumbo 9.5k.  Everything is set to max. 

I thought that the 'system' would auto-negotiate to get the best out of transfer speeds.  Are you saying I need to change the 323 setting to 7k and change the portables setting to 7k to get better transfer speed with the NICs? 

Ta in advance
Biscotte

OK I think I have understood.  As usual networks only goes as fast as the bottleneck - the rest negotiates up to and around this bottleneck speed. 
My switch is a Netgear GS108.  It has Jumbo frame support according to the manufacturer of 9,720 byte.  So it looks like this will not slow things down at all. 

You don’t say what small files are.  I assume .bat files say 150 bytes and other small stuff of under say half a meg.  I seem to move lots of files between 2 meg and 350 meg.  So again don’t think I will be slowed much.

Oh and the answer to the speed question is 8 meg per second (same as tops for 100Mbps) for some connections and 11-13 meg for other connections. 

BTW.  My 24 port 3com 100Mbps switch went phut.  As an interim measure I have been using the Netgear GS108 - with an old (read very old) 3com 10Mbps 24 port for printers and connections for those members of the family that are not being nice to me (!). 

Learning for me is as follows - my next purchase will be a 3com or similar 1,000Mbps 24 port switch, BUT now I will ensure I get one with big (9k) jumbo frame support. 

Thanks all
Biscotte

Soprano,
thanks for the comment, but been there, done that.  My machine is not that many feet from the DNS boxes - all cat6 between me and them.  Some of the rest of the house is cat5e.  And yes changed the cables after i had one go dud on me - Fordem is right, they don't last forever. 
Ta again
Biscotte