Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Hannes123]

Hi all!
After upgrading my network switch from 100Mbit to 1Gbit the DNS-323 transfers files pretty slow via SMB to my windows machine. With my old switch I got transfer rates round about 8Mb/sec from the NAS to the windows machine. Switching the NAS to GBit-Mode transfer rates go down to 3MB/s from the NAS to the PC via Samba.
The PC still works with an old 100Mbit network adapter. I expected the transfer rates to stay the same as with the old 100Mbit switch. Why does it decrease as soon as I use a new Gbit-Switch?
Using FTP transfer rates go back to normal values (8-9MB/s).

Any Ideas where to look for the problem?

Thanks for your help!

Hannes

[gunrunnerjohn]

Well, the switch seems the obvious issue.  Since the DNS-323 and the computer didn't change, that's the new component.

[ECF]

Do you get a GB link LED on all devices when you look at the switch?

[gunrunnerjohn]

Do you get a GB link LED on all devices when you look at the switch?

Since he stated that he still had a 100mbit NIC in the PC, I'd have to say no.

[fordem]

Well, the switch seems the obvious issue.  Since the DNS-323 and the computer didn't change, that's the new component.

Not necessarily - there is a cable between the DNS-323 & the gigabit switch that was running at 100mbps and is now running at 1gbps - I would replace this before replacing the switch.

Almost every instance I've seen where introducing a gigabit switch changed the throughput in a negative fashion, the problem has been a cable that is not up to scratch - 10/100 uses two pairs in a four pair cable, gigabit uses all four - 10/100 will run reasonably well with split pairs (provided the split is with the unused pairs), gigabit will not.

So - Hannes123 - please check your cables, make sure they are CAT5e or better and if they are hand assembled, make sure they are properly crimped.

[gunrunnerjohn]

Good point on the other cable.

[Hannes123]

Hi!
As soon as the NAS is connected the switch indicates an Gbit-Connection. Of course (as mentioned by gunrunnerjohn) this is not the case for the Notebook.
As suggested I checked my cables. I tried to connect the NAS to the switch with nearly every cable I found. No change at all! The cabels used are all cat5E. 


 

[fordem]

As suggested I checked my cables. I tried to connect the NAS to the switch with nearly every cable I found. No change at all! The cabels used are all cat5E. 

Are the cables purchased or home made?

[Hannes123]

All the cables are purchased! The cable used to connect the NAS to the switch ist the one that came with the NAS. The cable from the switch to the notebook seems to be from "belkin".

[Tank_Killer]

It might be possible you have a lossy switch,  however I have a few questions.  Where is your router in this equation? are any devices on said router? or is it just WAN and uplink to switch?  Reason I ask is perhaps you might have a loop in your network (I.E. the switch is plugged into itself, or you have 2 uplink cables running from your router to new switch) This can be a caused for poor performace.

try to disco everything on your switch except the NAS and your PC to eliminate any other cabling issues, as fordem suggested I would make those cables for this test brand new. for giggles.

gl

TK

[gunrunnerjohn]

I really don't think that the changes of multiple factory manufactured cables failing are great enough to go to the trouble of actually making cables.  There is clearly something else going on here.

It could be a defective switch, I've sent back two SOHO gigabit switches within the last 6 months for defects, (A netgear and a Trendnet).

Try different ports on the switches, many times a specific port will fail.

[fordem]

I really don't think that the changes of multiple factory manufactured cables failing are great enough to go to the trouble of actually making cables.  There is clearly something else going on here.

Agreed - whilst cables do cause a number of the failures in gigabit installations - in this case multiple factory manufactured cables have been tried - the only thing left is the switch.

[Hannes123]

Where is your router in this equation?

The router is just used for WAN uplink to the new switch and as DHCP-server.

Reason I ask is perhaps you might have a loop in your network

I checked this twice. No loops!

There is clearly something else going on here.

I used WireShark to record 100.000 Frames during several copy processes from the NAS to the Notebook.
When the NAS is set to 100Mbit there seem to be now errors in the communication between the devices.
As soon as I switch the NAS to 1Gbit many retransmissions of packages occure due to inconsistencies in the sequenznumbers. I guess this causes the low transfer rate.

May this be caused by what you call a "lossy switch"?

[gunrunnerjohn]

I'm betting on the switch, of course there could be some flaw in the DNS-323 network adapter as well.  I have run my DNS-323 and DNS-321 at gigabit with jumbo frames for some time, and I haven't seen any networking issues.

I DID have the speeds suddenly fall way down, and it turned out the switch had suddenly forgotten how to handle jumbo frames when it got warm!  I also had a total failure with another switch, a couple of the ports on the switch died.

I'd have to try swapping ports on the switch first, then the whole switch next.

[fordem]

When the NAS is set to 100Mbit there seem to be now errors in the communication between the devices.
As soon as I switch the NAS to 1Gbit many retransmissions of packages occure due to inconsistencies in the sequenznumbers. I guess this causes the low transfer rate.

Retransmissions will certainly cause low throughput - but retransmissions due to "inconsistencies in the sequence numbers" - that I've never heard of before.

In large - let's  call them "multi-path" - networks, networks (such as the internet) where there are multiple possible paths that packets can take along their trip from point A to point B, it is entirely possible for packets to arrive "out of sequence", and it is the function of the network stack to reassemble them in the correct sequence - this in fact is the reason that the sequence numbers exist - but on a local LAN where there is only one path between source & destination, out of sequence packets should not occur (assuming of course, that that is what is meant be "inconsistencies in the sequence numbers"), and the stack should reassemble the packets rather than reject them or request retransmits.

Sounds to me like some sort of corruption is occuring somewhere - do you have a second windows system to test from?  Is there reduced throughput there also?