I also get around 6MB/s when using the Alix 2c3 as a simple router. Problem seems to be the combination...
vr2 -> vlan7 (vlandev vr2) -> pppoe0 (dev vlan7) I also don't thing the Alix is too slow. As i said...50.000kbit/s down via pppoe0 works w/o issues. On Fri, Oct 10, 2014 at 01:10:39PM +0200, Stefan Sperling wrote: > On Fri, Oct 10, 2014 at 12:23:36PM +0200, Mark Patruck wrote: > > I'm running 5.6-current on a Alix 2c3. The box is connected > > via pppoe(4) and VDSL 50Mbit down/10Mbit up - max-mss is set > > to 1440. > > > > Running a few speed tests, i get almost always > 50.000kbit/s > > down, but not more than 400-600kbit/s up. > > > > Just for testing purposes, i started httpd(8) and tried to > > download a 1MB test file over the internet from another machine. > > > > $ ftp http://1.2.3.4/test1MB > > Trying 1.2.3.4... > > Requesting http://1.2.3.4/test1MB > > > > After about 8 seconds it shows 128KB, then...few seconds later... > > --stalled--....few seconds later 256KB....--stalled--.... > > > > 65 seconds later, the download has finished. > > > > The same configuration (freshly installed OpenBSD 5.6-current) on > > another Alix 2c3 shows exactly the same issues...download fine, > > upload < 600kbit/s. > > > > Just to make sure there is nothing wrong with cabling, VDSL modem, > > i tried the same configuration on an older Celeron laptop with > > ale(4) nic...no issues at all. I get around 8.000kbit/s. > > > > Any clues? (vr(4) issues?) > > I don't think vr(4) is your problem. > From a net5501 soekris (similar hardware) I can download 6 megabytes > per second of a file on the soekris' hard disk via a LAN-facing vr(4) > interface, served over HTTP with nginx (on 5.6-stable). > > You could run measurements with tcpbench(1) to rule out problems > at the network/driver layer. In my testing an Alix.2d2 lx800 (running > 5.6-stable too) is slightly faster with tcpbench (Avg Mbps: 92.490) > than the net5501 (Avg Mbps: 86.949), both using vr(4) interfaces > connected to a gigabit switch. > > Perhaps it's worth mentioning that the vr(4) interfaces are part > of a bridge(4). I'm not sure if that affects throughput but if > it does plain vr(4) interfaces could be faster. > -- Mark Patruck ( mark at wrapped.cx ) GPG key 0xF2865E51 / 187F F6D3 EE04 1DCE 1C74 F644 0D3C F66F F286 5E51 http://www.wrapped.cx