SG-1000 oddity (maybe hardware bug?)
-
I suspect there might be a hardware bug in the SG-1000, or at least in the unit I received. Some background…
Recently, I deployed a new SG-1000 in my parents' house to replace their aging PC-Engines ALIX, since it was not fast enough to handle their newly upgraded Internet connection (100/35), and since it would no longer be supported in 2.4.
Before installing it, I configured and tested it on my own network and found that it should perform just fine, topping out around 150 Mbits/sec, which is more than sufficient. After installing it, however, I was having some significant issues, with tests showing throughput only around 20 Mbits down, but still 30+ Mbits up. Initial investigation showed that it wasn't the SG-1000 getting bogged down on CPU or anything else obvious. I suspected the issue might be with the ISP, but connecting my laptop directly to the cable modem showed that their connection was getting the advertised speed.
At this point, my (admittedly rusty) networking knowledge suggested to me that the performance degradation in one direction only might be due to a duplex mismatch. According to the Interfaces status, everything was as I expected it should be, cpsw0 showing 1000base-T full-duplex to the modem, and cpsw1 showing 100baseTX full-duplex to the 100 Mbit switch that everything in their house was connected to. Just to verify, I swapped the port assignments and saw the same issue. I then swapped it back and tried plugging my laptop directly into cpsw1 and everything seemed to be fine. So then I swapped out the switch with a new gigabit switch, and everything is now working great.
So... In the end, everything is working out just fine for my install, but I suspect there might be a problem with the built-in switch that's part of the cpsw chipset, when the two ports are running at different speeds. Obviously I can't definitively prove this suspicion since I only have the one unit, but I figured I'd put this out there in case someone else might be experiencing the same issues, or if maybe Netgate could test and confirm this potential bug.
-
One other thing I forgot to mention…
The performance problem I was having before replacing the switch was only present on TCP, not UDP. Google-operated sites like YouTube (which uses Google's new QUIC protocol) still performed as expected. This, I believe, is more evidence toward my assessment that it's a duplex mismatch, as QUIC doesn't require much bi-directional traffic, where normal HTTP requires all of the TCP ACKs going back.
-
Please contact our support https://customercare.netgate.com/
-
What was the result of this. I have a client that is having this exact same issue. As soon as they use the 100MB switch their speeds drop to nothing.
-
What was the result of this. I have a client that is having this exact same issue. As soon as they use the 100MB switch their speeds drop to nothing.
I was informed by our support staff that they are working with you on this.
-
I worked with GordC and discovered that it was ISP related issues. If you are having issues please don't hesitate to reach out to us.
-
I am still working on this. At this time I am not completely sure what is going on. I am going to be setting this firewall up in my lab and run some tests.
