Strange problem with speed of OpenVPN tunnel

skritland

Hello everybody!

I have very strange problem with connection speed through VPN. I have VPN server installed on VPS OpenVZ Debian server with fast connection (100Mb/s). VPN client is Windows XP computer located in China and connected to internet by medium speed connection. Strange thing is that when I'm downloading a file using VPN and web browser, speed is approx 20-40 kB/s. When I'm using download manager that divides file into parts and downloads them simultaneously, speed is 150 - 300 kB/s (each part is downloaded with speed 20-40 kB/s). I've checked my server: the same file downloaded as one part (using wget) to the computer that is VPN server can reach 5 MB/s - so it is not server connection speed problem. It can't also be problem with connection VPN server-client - data is going as one compressed and encrypted UDP stream so it shouldn't be difference between downloading in one or many parts. The same is with streamed videos - it is to slow to watch it without waiting for buffering but when I open several tabs with different videos all are loading with the same speed as if only one was opened. I've been trying turning off encryption/compression but without any result.

I have no idea if it is OpenVPN problem or maybe Linux server or WIndows client configuration. Maybe someone has some idea what could be wrong. Below config files for server and client:

server:

port 1194
proto udp
dev tun
ca keys/ca.crt
cert keys/aaa.crt
key keys/aaa.key
dh keys/dh2048.pem
tls-auth keys/ta.key 0
server 10.19.87.0 255.255.255.0
ifconfig-pool-persist ipp.txt
push "redirect-gateway def1"
push "dhcp-option DNS DNS1_ADDRESS"
push "dhcp-option DNS DNS2_ADDRESS"
cipher BF-CBC
comp-lzo
user openvpn
group openvpn
persist-key
persist-tun
status openvpn-status.log
verb 4

client:

client
dev tun
remote SERVER_ADDRESS.com 1194
proto udp
resolv-retry infinite
nobind
persist-key
persist-tun
ca keys/ca.crt
cert keys/bbb.crt
key keys/bbb.key
tls-auth keys/ta.key 1
ns-cert-type server
cipher BF-CBC
comp-lzo
verb 4

janjust

this depends a bit on your connection - try tweaking the following parameters

fragment
mssfix
tun-mtu

in UDP mode I'd start out with

fragment 1200
mssfix

In TCP mode I'd start with

tun-mtu 1400

this setting is needed on both client and server.

skritland

I've tried this but it didn't help. I've thought it must be like that and I've used this slow connection. But few days ago I've tried ssh tunnel using Putty as local proxy and it is faster 2-3 times than OpenVPN. Maybe someone knows why that happens?

mdylewski

Hi, any solution? I've got the same problem :/

dgcruzing

I would like to know a few tweaks too as been running some tests and pptp v openvpn speeds to the same servers are deadly slow..

Mine is Android based going out of China to UK/Sing/US/Germany/Swiss servers and the pptp is by far better for any sort of streaming..

So any tweaks that could be made would be appreciated..

Sent from my Nexus 10 using Tapatalk

Laakkus

Test with a Linux client if possible. I have similar problems with Windows clients, the same time Linux client speed is great..

Lähetetty minun GT-I9300 laitteesta Tapatalkilla

Xottabych_ua

I've find something useful on this.
I have 2 OpenVPN servers with 3 clients connecting to each. Actually, it is site-to-site scheme.
Both servers are VPS with same capabilities and from the same provider, but in different DC's. Servers configured as default without any tuning with mtu,mss, etc.
Clients are linux (centos) virtual machines on proxmox virtualization system.

iperf test from client to first server shows about 90Mbit/s without tunnel and 3 Mbit/s trough tunnel. Zats crazy.
iperf test with second server shows

[ 4] local 162.243.x.x port 5001 connected with 92.60.x.x port 60654
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-10.2 sec 31.5 MBytes 25.8 Mbits/sec
[ 5] local 172.17.24.1 port 5001 connected with 172.17.24.2 port 39695
[ ID] Interval Transfer Bandwidth
[ 5] 0.0-10.2 sec 21.5 MBytes 17.6 Mbits/sec

The difference is only in OpenVPN configuration:
On the first server it is one instance with one tun0 interface for all clients.
On the second - three instances with different ports and tun's, and clients are configured to connect to different ports accordingly.
I cant find anything else can cause such difference. Hope this helps somebody

P.S. One more difference that first server is Ubuntu server and second one is Centos 6.5.
But can this be the reason? It is not bad idea to test same configs on different kernels...