Low speed in cascade (latest stable: c0c1b91 )

[nickdsl]

Hello!
I discovered low speed in cascades of SeVPN.
My tests:
iperf3 between external (white) IP's
iperf3 between internal (grey) IP's

What i've got:
bandwidth between peers (external) was 2-5 times faster than bandwidth in the cascade between same peers.

Solution:
downrade to commit 4b65e25 ( v4.20-9608-rtm -> v4.18-9570-rtm )

Question: what's wrong with v4.18-9570-rtm ?

[thisjun]

Can you explain your VPN and network configuration?

[nickdsl]

thisjun wrote:
> Can you explain your VPN and network configuration?

I have multiple XenServers.
Each server has own Single-Server Private Network named: vpn-internal
Every VM on server has virtual ethernet interface connected to vpn-internal
On each XenServer runs one VM with installed softether vpn server.

(vm runs on: OpenSuse 13.2 Linux)
Interface configuration of VM with vpn server role:
eth0: (white ip) (looking directly to internet)

eth1: 10.20.30.x/24 (local ip address of current machine)
eth2: without ip (local bridge to VIRTUAL HUB)

I use two interfaces because i found one bug:
if i make local bridge on eth1 interface, then i cannot ping this host from other hosts


On one of VM with sevpn i create cascade connection to other VM.
But it is not all. I use this script at startup:

cat /etc/init.d/md-eth
#!/bin/bash
#test script for ethtool
### BEGIN INIT INFO
# Provides: md-eth
# Required-Start: $network
# Required-Stop: $network
# Default-Start: 3
# Default-Stop: 0 1 2 6
# Description: MD-eth
### END INIT INFO

interfaces="eth0 eth1 eth2"
case "$1" in
start)
echo -n "Start md-eth..."
for i in $interfaces
do
ethtool -K $i rx off tx off sg off tso off gso off
done
echo "done"
;;
stop)
echo -n "Stop ... done"
;;
restart)
echo "Restart md-eth"
$0 stop
$0 start
;;
status)
echo "************* status md-eth******************"
for i in $interfaces
do
echo "************* interface $i *********************"
ethtool -k $i
done
echo "************** end status **********************"
;;
*)
echo "Usage $0 {start|stop|restart|status}"
esac

Without this script speed will be very very low (Kbit/s). After that speed will grow up.
But it is much lower than speed between white ip's.


P.S. sorry for my english)

[nickdsl]

Here is my schema:
i have two dedicated servers with XenServer installed.

I tested on local virtual stand. I created same schema in VMWare environment.

[nickdsl]

TESTS (iperf3):
XEN1 -> XEN2:
10.113.80.201 -> 10.113.80.202
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.04 sec 578 MBytes 483 Mbits/sec 0 sender
[ 4] 0.00-10.04 sec 578 MBytes 483 Mbits/sec receiver

X2 -> X1:
10.113.80.203 -> 10.113.80.204:
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 139 MBytes 116 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 138 MBytes 116 Mbits/sec receiver

X2 -> X1:
192.168.1.2 -> 192.168.1.1:
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 125 MBytes 105 Mbits/sec 123 sender
[ 4] 0.00-10.00 sec 125 MBytes 105 Mbits/sec receiver




TESTS (iperf3):
XEN1 <- XEN2:
10.113.80.201 <- 10.113.80.202
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 735 MBytes 616 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 735 MBytes 616 Mbits/sec receiver

X2 <- X1:
10.113.80.203 <- 10.113.80.204:
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 142 MBytes 119 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 141 MBytes 118 Mbits/sec receiver

X2 <- X1:
192.168.1.2 <- 192.168.1.1:
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 126 MBytes 105 Mbits/sec 52 sender
[ 4] 0.00-10.00 sec 125 MBytes 105 Mbits/sec receiver

[nickdsl]

on local stand speed between x1 and x2 is same
but on production speed in cascade is lower down to 2-8 times.

i have not any idea why it happens.

[thisjun]

I think we should isolate the problem whether the cause is localbridge or cascade.
Please try that a client connect to the VPN server and measure the speed between the client and the localbridge or the client and the cascade .