I was wondering about the general view on mikrotik’s openvpn solution. I understand that some people hate it because of the lack of udp and proper site-to-site. As for me, I need tcp (some proxy clients) in a manyclient-to-server setup.
The initial tests (server on gigabit line, with 10 clients of various internet connections 10-120Mbps) showed a proper latency, and that the download from server side was 2-3x faster (20Mbps->50Mbps) compared to previous linux setups. However, the upload to the server in most of the cases couldnt go beyond 5Mbps.
Are you sure that it was not because of the asymmetric nature of the subscriber line that the clients used?
All in all I have good experience with MT VPN solutions including openVPN since version 5.x when it became stable for the first time.
However under some circumstances vpn seems to become unreasonably slow. For example server to server speed test shows the full 100mbps, but the VPN performance (openVPN, pptp etc…) would not go beyond 1-2mbps no matter what I do. The same setup on a different network might perform much better … like 60x better. So what I’m trying to say is that you may not have control over all the factors that detemine the actual performance of the VPN.
Yes, I am quite sure it is not due to the asymmetric line: for example my own internet connection is 120/10Mbit, and I am unable to reach more than 50/5. Compared to this, on linux based openvpn I was able to reach 20/8. So as I said, more download speed, less upload.
Hope I will not find that unreasonable behaviour. Maybe it is with some versions? Or MTU?
I don’t always have performance issues. Some installations work fine and I can use 60 out of 100mbps, sometimes only 5 or 10. I thought it was a bandwidth problem to the server, but the servers can talk to each other 100/100 when I do a bandwidth test.
It’s not an openvpn issue, but a generic vpn issue that I seem to have with ROS. There seem to be some invisible network parameters that can slow VPN.