Upgrade to 2.5 GbE over USB worked... but there is a problem...

[TML59]

(I also posted this on other forums but -to date- wasn't able to solve my problem)

In short: I purchased a UGREEN USB-to-2.5Gbe converter, followed instructions on the Internet and was able to install drivers and got the NAS (DSM 7.1) to report a 2500mbps Ethernet connection. My cables are Cat7, my switches 2.5 GbE and so is my Windows desktop Ethernet interface. So all seems fine.
When I measure file transfer speed however, I'm still seeing 70-80Mb/s (= 1GbE). An iPerf3 test reported 133Mb/s (which is low for 2.5 GbE but above 1GbE, indicating that there is a 2.5 GbE capable connection, so it seems that 'hardware-wise' things are correct).
Why am I still seeing a 1GbE network speed then??
Any suggestions?

 

[strikes2k]

I have an RS2423RP+ with 10 gig networking. I have 10 gig from my desktop all the way to my Synology. I should see blazing speeds. My Windows desktop has a couple of drives in it - An NVMe drive for boot and an old 3 TB WD spinning disk for storage. When using the spinning disk I see anything from 70 MB/sec up to about 160. It varies I'm guessing depending on what else the drive is doing. But if I copy that same file to the NVMe and then do the copy from the NVMe I'll see 400+ MB/s. So I've come to the conclusion that my bottleneck is the spinning disk. What type of drive are you copying from?

 

[TML59]

The drives in my desktop are NVME, those in my NAS 'spinning disks'

 

[Rusty]

Why am I still seeing a 1GbE network speed then??

Are you sure that you are communicating on the IP address for that 2.5G NIC and not on your 1G? Are you connected via IP address or are you using a netbios name? If so, it will default to 1G.

So try and make a connection from your computer to your NAS using an SMB connection targeting the explicit IP address of the 2.5G NIC. Also, you can just disconnect the 1G NIC cable, leaving 2.5G connected and then try and access it using its netbios name, it will have to use 2.5G.

If tests in that case also terminate under 200MB/s then you can say "you have an issue".

What type of drives and array configuration is inside your NAS? What model of NAS is it?

 

[TML59]

Thanks, Model NAS is in my profile above the posts: DS216J. Connection is ONLY via the USB>2.5GbE converter (I removed the standard connection to the 1GbE bus). I connect via a (fixed) IP-address.
Not sure how to make a connection via SMB... if you can explain (maybe the iPerf3 test does that?)
HD's are WD's, one 3 Gb (WD30EFRX) one 4 Gb (WD40EFZX)

 

[Rusty]

Well the 216 and on top of that J will indicate a possible mirror with those 2x3TB drives. So getting anything more then 200MB/s will be pushing it due to specs of the J model itself. Still, going over 100MB/s should work but don't expect miracles.

if you can explain

If the 2.5G connection is the only one then you do not have to do anything special here considering the NAS is only accessible on that particular interface.

With all said and the fact that even iperf can't go over 133MB/s. I would say there is not much you can do.

Have you tested with a single large file? Are those speeds you are getting from that kind of test?

 

[TML59]

I'd be happy with 150Mb/s... I tested with a 10Gb file, no differences.

 

[Coop777]

spindle drives will be a bottleneck, especially if they are slower drives.

 

[TML59]

I know that, but if an iPerf3 test shows 133Mb/s, to me that means that I should be able to achieve that speed also when I do a normal file-transfer. And an rotating HD should be able to achieve that speed.

 

[strikes2k]

I know that, but if an iPerf3 test shows 133Mb/s, to me that means that I should be able to achieve that speed also when I do a normal file-transfer. And an rotating HD should be able to achieve that speed

Isn't iPerf a network monitoring tool? You can't compare disk copy speed based purely on the raw network speed. If anything, that is telling you that yes your network is capable of the speeds it indicates. Storage performance has other factors as I tried to convey in my first post in this thread.

 

[TML59]

OK, I believe that. But what remains is that I had a transfer speed of about 77Mb/s before I made the changes (when I used the original 1 GbE Ethernet connection on my NAS). Now I connect via 2.5 GbE Ethernet and the transfer speed is unchanged.
If the raw network speed is faster now and the overhead is unchanged, shouldn't I expect (much) more than exactly the same 77Mb/s after my upgrade then?
Something is not right but I can't find what it is. I'll try to do some other tests in the coming weeks to eliminate possible root causes. In the meantime however, I continue to appreciate any additional suggestions and tweaks I could make.

 

[Rusty]

Here’s an idea. Try and push multiple copy tasks from different devices. See what the cumulative total speed you can get and just how much you can saturate the network and the drives. Imho, that J should be able to push about 100MB/s but anything more that might be pushing it.

Just an idea.

 

[fredbert]

I would be inclined to connect the 1GbE port direct to Mac/PC and then monitor the activity that's happening when the LAN is connected to the 2.5GbE port. Either using the Resource Monitor or SSH and things like top.

iPerf3 is just a network performance tool, it doesn't test the disk read/write. But 133 MB/s still seems low (I assume you meant MB/s not Mb/s) for a 2.5GbE test. Only when this is is around 300 MB/s should you look at the data transfer speeds of file sharing.

You don't have any traffic control configured? And the 2.5GbE port is assigned as the default/primary is Service Order?

 

[TML59]

Here’s an idea. Try and push multiple copy tasks from different devices. See what the cumulative total speed you can get and just how much you can saturate the network and the drives. Imho, that J should be able to push about 100MB/s but anything more that might be pushing it.

Just an idea.

Thanks, I can do that but at this moment in time I don't have another device available regretfully. Hope that will change in a couple of weeks

I would be inclined to connect the 1GbE port direct to Mac/PC and then monitor the activity that's happening when the LAN is connected to the 2.5GbE port. Either using the Resource Monitor or SSH and things like top.

iPerf3 is just a network performance tool, it doesn't test the disk read/write. But 133 MB/s still seems low (I assume you meant MB/s not Mb/s) for a 2.5GbE test. Only when this is is around 300 MB/s should you look at the data transfer speeds of file sharing.

You don't have any traffic control configured? And the 2.5GbE port is assigned as the default/primary is Service Order?

Thanks, let's start with the last line. ;-)
Traffic control: Not AFAIK, what can I do to check that?
2.5 GbE port assignment: Can you explain? Do you mean on the NAS side or the PC side? How can I check that?

 

[fredbert]

Traffic control: Not AFAIK, what can I do to check that?

2.5 GbE port assignment: Can you explain? Do you mean on the NAS side or the PC side? How can I check that?

Use Service Order to access a pop-up window to order the priority of your interfaces: top of the list is highest priority. I have Bond 1 as highest and then LAN 4. The displayed list order in Network Interfaces doesn't necessarily reflect the service order.

Also in your 2.5GbE Network Interface settings, under IPv4 have you:, and not for the 1 GbE port

If you have SMB multi-channel enabled under SMB advanced settings, switch it off ... at least for now.

 

[TML59]

Thanks: nothing in traffic control
Service order: as I have only one active connection (plus the inactive 1GbE connection) I can't change the order
Default gateway: In my router I defined fixed IP addresses for the 1GbE and 2.5 GbE connections. In the NAS under IPv4 the settings I see "Get network configuration automatically (DCP)" enabled and therefore I can't select anything else

 

[TML59]

Update: I read some more info on the Internet and enabled SMB3. That increase my file transfer speed from 77 MB/s to 100 MB/s. Still too low (and maybe I would have seen the same on the 1GbE connection) but a significant improvement