Hey all,
I have an XGS-PON 8/8Gbit connection via my ISP, using PPPoE through an SFP+ module in my CCR2116.
The best I can seem to get in throughput is 2.5 Gbit/s. I know the connection can do more, but I can’t figure out what the bottleneck is on the network.
If I bypass my Mikrotik completely I get substantially higher speeds.
I know doing multi-gig isn’t super common place yet, especially on PPPoE, but I’m hoping someone out there knows what might be the cause of this.
Did you solve this? I have a similar problem with a pppoe connection, only the link should be 1G but i can squeeze down only 400M
Bottleneck is probably because PPPoE is a single-threaded process stuck on 1 CPU-core (out of 16) when it comes to encapsulation/decapsulation and therefore you are probably hitting its limit.
Take a look at the port speed. 2,5Gbps is a classical option. Maybe your XGS-PON is synchronized at 2,5Gbps?
how much is your throughput without PPPoE ?
No, I never did. I’m still stuck at 2.5ish Gbps. It seems to be because PPPoE is single threaded on RouterOS, and as a result only a single core is used and it hits 100%.
No, this is definitely not it. It’s PPPoE slamming out only a single core.
Pretty close to line speed.
I picked up a TP-Link router that can easily max out the connection because it does hardware offloading of PPPoE. Unfortunately it’s a TP-Link router so it doesn’t do what I need it to do.
So why on my RB5009 PPPoE getting 2.0Gbit/s (full speed) with just 20% CPU load and almost all cores are equal?
http://forum.mikrotik.com/t/pppoe-client-on-xgs-pon/181049/1
The bottleneck is likely due to the PPPoE process being CPU-intensive, especially at multi-gigabit speeds. The CCR2116, while powerful, may still struggle with PPPoE at such high throughputs due to single-threaded limitations of the PPPoE process in RouterOS.
To improve performance, try the following:
If these steps don’t resolve the issue, the CCR2116 may not be ideal for handling multi-gigabit PPPoE connections, and a more specialized solution might be necessary.
PPPoE is horrible unfortunally and RouterOS has no hardware-offload for it. It’s strange some ISPs hold on to 20 year old concepts.
I guess it suits them well for a few purposes … one of them is user management (less fuss to e.g. assign static IP address and IPv6 prefix). And obviously they don’t bother about (under)performance of 3rd party routers, they just care about performance of “their” equipment (ISP-provided CPEs/routers and certainly their PPPoE servers).
I guess no new ISP starts their network with PPPoE … but for legacy ISPs it might be a big step to ditch it (a huge disturbance in their operations for dubious gains).
I run an ISP and built out DHCP to move away from PPPoE and I gotta say I lose a lot by moving to DHCP. One issue is that I no longer have a nice virtual interface for every customer that’s maintained by LCP. This may not sound like a big deal, but in the DHCP world and with firewalls blocking ICMP, I have no way of knowing if the customer is online or not shy of arp ping.
Also, if something happens and my customers go offline for a minute, I’d never see it, but with pppoe I see everybody reconnect. Lastly, if I have a lot of locations I have to create huge flat network which is ugly or I have to partition my IP space between the locations.
With pppoe everything is /32 so I simply inject the /32 routes into my routing protocol and everything works, with DHCP I have to use option 121 and /32 which doesn’t work everywhere, or go through the pain of making a subnet for each site.
Routing /30 networks is pain and I don’t immediately recover unused addresses unless I run a very short lease which has it’s own issues.
About the only con to pppoe is the lost 8 bytes, and performance issues past 1G, everything else works better.