If I have a 10Gbps WAN connection, does that mean I should be able to fully utilize 2.5Gbps on each of the 3 LAN ports, or is the total throughput across all LAN ports limited to 2.5Gbps?
Yes, if you have three 2.5 Gbps LAN links, you can use 7.5 Gbps across them. It’s more typical to connect a 10 Gbps-capable switch to the LAN4 port, but either scenario will work.
1 Like
With a 10 Gbps uplink, I’m only able to achieve a maximum of 2.5 Gbps across all LAN ports. This means that if I run a speed test on LAN2, for example, the speed on LAN3 will be limited to 2.5 Gbps. If I disable hardware acceleration, I can push speeds above 4 Gbps(Only tested on 2 device), but this results in 50% system load. For WAN2, running speed tests usually caps most servers at 1 Gbps, with only one or two servers allowing me to reach the full 10 Gbps. However, when I switch LAN4 to WAN, I’m able to hit 10 Gbps on all speed test servers. Is this potentially a hardware issue? I’ve tested all LAN ports, and as long as the connection uses the WAN2 SFP port, most speed test servers are capped at 1 Gbps, although torrents work fine.
Shouldn’t be a hardware issue, other than the chipset may Can you describe more how you are running the tests?
Some things you can try are to disable flow and flow control:
ssdk_sh flow status set 0
ssdk_sh port flowCtrl set 5 disable
ssdk_sh port flowCtrl set 6 disable
Better speed, how to set it as persistent?
Does set 5 and set 6 only disable flow control on certain ports? I want to do some playing around with this please. I am using the first 2.5Gb port from the left, and the first SFP+ port from the left.
Does flow status set 0 disable hardware offloading by hand? that command causes a lot more CPU load on speed tests.
Thank you Jeff.
You can add the commands to a shell script /cfg/post-cfg.sh to get the commands to execute after every configuration is applied.
Port 5 is WAN2 and Port 6 is LAN4. Yes, disabling flow will disable hardware acceleration as well, and there may be corner cases where this results in increased speed, so we’re looking into how to improve this.
Thank you Jeff. Are you willing to share with us the “port” vs physical location on the front of Route10?
Sure.
WAN1: ssdk port 4
LAN1: ssdk port 3
LAN2: ssdk port 2
LAN3: ssdk port 1
LAN4: ssdk port 6
WAN2: ssdk port 5
It’s this way because it’s best for the HW layout.
1 Like
Thank you so much Jeff!
We are still working to understand this issue and find a resolution. However it’s difficult to pick out the actual findings in your tests, and what your tests represent.
In order to best determine what to do, let’s start with a basic testing methodology.
Please test the Route10 on a new cloud managed site with no other connected devices. Use only your wired laptop or workstation pc to begin.
Saturate LAN1, for example, with a 2.5G test. Now confirm this result is identical on the other LAN ports in sequence.
Next, we can try adding a second wired client to a second LAN interface simultaneously. Now saturate the 2.5G on L1, then initiate another test across L4. Please shuffle these combinations around and test as many as you see fit, but make sure that your tests can be repeated consistently.
If I read your message correctly, you are seeing correct limitations at 2.5G. I believe this is to be expected, but please help me understand what your expectations are instead.
Before proceeding further with your other configurations, such as disabling flow control, I’d like to see where thing stand at this point first. Could you return with the results soon?
Could I ask what your tests and findings were as well? Did you see unusual results that match the descriptions in this thread?
@jaianna I grow tired of repeating myself. Abuse will not be tolerated on our forums, whether you are abusing us, our products, or any of our community members. You have no evidence that we’ve created false accounts to support our team or product, and I assert to you and anyone else interested that we have not done so. The product stands on its own and you’ve seen that we have plenty of faithful community members willing to help improve the product in responsible ways.
If you do have exploits, for which you have no evidence that you actually do, I would ask that you divulge them in an ethical manner, where you will be handsomely rewarded for your efforts. Handling the situation in any other way only accentuates the suspiciousness with which you have handled everything up to this point.