January 20, 2015 at 3:25 pm #44165
1st site has 2 DSL connections:
10/10 Mbit/s SHDSL with latency of 2-3ms
16/1 Mbit/s ADSL2+ with latency of 12-18ms
The other site has 300/300 Mbit/s ethernet connection with low latency
I have configured 2 extra VPN tunnels with bonding as on this example:
Although I am using UDP tunnels and no compression and use PSK instead X.509.
Well, everything worked, that is clear. But what was unexpected – When using bonded VPN connections with load balancing, I only get about 5 Mbit/s speed. When I use failover only, then i get 10 MBit/s or 14 Mbit/s, depending which tunnel i select as primary.
I looked back at older posts that others have had that problem too, but no real solution.
I guess problem might be in different latency in DSL connections causing out-of-order TCP packets that keeps data connection speed low.
I am sure both DSL connections were used, cause lights on both modems were blinking. Also I can see that from MRTG graphs. Also when pinging in VPN tunnel the reply alternated between 4ms and 12+ms.
If anyone has any suggestions, before I tear down this bonding connection? Or I can leave it and use it just for failover. And it is good, that at least i can use SHDSL for upload and ADSL for download (asymmetric routing using different VPN tunnels).March 28, 2015 at 1:32 pm #53655
Old topic : )
Well, my situation is a bit different : 2 standards xdsl (2 X 2.5Mb) -> zeroshellserver with QoS & Netbalancing-> LAN / No VPN
When I connect the LAN directly to one of the xdsl (bypassing physically the zeroshellserver), I only have 2.5Mb, that allright. In this configuration there are no lag at all and the xdsl box monitoring shows normal activities.
But there are similarities with the problem you got when going through zeroshell :
1/ I expect my load threshold to be 5Mb (both xdsl full capacity) ok, but it’s almost always at 100% – in zeroshell monitoring – and everyone lags.
2/ I still have monitoring on each xdsl through the admin of my internet provider’s boxes : When zeroshell indicates it’s already at 100% of the expected 5Mb, the monitoring of the boxes show that they’re not even at 20% of their capacities (1.5Mb/s total consumption)..
3/ I haven’t try to switch to FAILOVER only, which I will, but there’s one thing which is suspicious to me :
– I have a static Ip adress which bypass the QoS rules (but not the netbalancing), and if i get synch problems sometimes (eg : with httpS), i DO NOT HAVE lag problems at all…
– The inner-LAN is connected to zeroshell server with a switch (2 ethernet RJ45 cables) and a network card that are both 10/10 only.
From the led-monitor of the switch, i read it is @ 100% usage already
So I wanna try two things :
– switch to failover
– replace switch and card for 10/100 devices
but I have the feeling it has connection with the QoS …
I’m not sure it’s of any help, but it’s the closest topic to my problem I’ve read so far…June 18, 2015 at 8:35 am #53656
Have been trawling the forum for people with similar issues..
I have 2x FTTC connections and third FTTC connection across a PTP Link all bonded.
When the 2 local FTTC connections are bonded with similar latency i get the correct throughput.
The problem comes when i bond the third (Slightly Higher Latency) conenction the performance drops down for single tcp transfers.
All very strange.
You must be logged in to reply to this topic.