Web sQuad ISP - Feedback Thread #2

@websquadza I don't like this L2TP tunnel for my static IP neh.

I had Claude help me diagnose it, and he says it's you guys :D
Some useful feedback from my end:
- Direct over Vuma I get about 1 Gbps down, through the tunnel about 290, and upload is 210 either way. (using fast.com)
- No router (as a reminder, MikroTik CCR1009-8G-1S-1S+) core went above about 37% during the tunneled test, and the tunnel work was spread across all of them.
- 1540-byte packets reach their tunnel server unfragmented, so the tunnel can carry full 1500-byte packets.

Seeing a drop off in speed when using the tunnel vs leaving it disabled, over NNTP too
 
@websquadza I don't like this L2TP tunnel for my static IP neh.

I had Claude help me diagnose it, and he says it's you guys :D
Some useful feedback from my end:
- Direct over Vuma I get about 1 Gbps down, through the tunnel about 290, and upload is 210 either way. (using fast.com)
- No router (as a reminder, MikroTik CCR1009-8G-1S-1S+) core went above about 37% during the tunneled test, and the tunnel work was spread across all of them.
- 1540-byte packets reach their tunnel server unfragmented, so the tunnel can carry full 1500-byte packets.

Seeing a drop off in speed when using the tunnel vs leaving it disabled, over NNTP too
Howdy

You by any chance have an arm based mik to compare against? CCR-1x tile processor has some weird internal bottlenecks (even when showing low usage). We saw similar when terminating Openserve LNS connections (the whole Openserve network is L2TP) on 1072 and 1036s. It’s definitely not a protocol / overhead issue. However you have a decent box, that just doesn’t like L2TP. For comparison a dual core arm box gives you between 450 and 500M and quad core will max out 1G with ease (they usually have faster clocks). I’m glad to hear you’re using the jumbo frame to get full frame on the LNS.

Drop me a PM so I can get your details.
 
Top
Sign up to the MyBroadband newsletter
X