Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Sebastian Moeller
To add to Joel's point, I can do my own catX cable runs and connect sockets/plugs to the cables, but I lack the tools for fiber-splicing... as cool as that would be it is going to be hard to justify multi-100s EUR for a splicer.. That still leaves short distance in the main computing area of an

Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Joel Wirāmu Pauling
The XG PON ONT units from Nokia/Huawei are coming with only 10G NbaseT (usually singular port) only in the consumer access space. No SFP+ We have rolled out XG PON on the PON side to 70% of the country here (NZ) over the last 2 years. Only a small % of that are actually making use of the XGPON on

Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Joel Wirāmu Pauling
Completely aside I have never got Cake SQM to work with connection's beyond about a gigabit biderectional ; without loosing gigabits of throughput even when running on beefy hardware. Has been a problem here for some time now. On Fri, 17 Dec 2021, 9:36 pm Joel Wirāmu Pauling, wrote: > The XG PO

Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Sebastian Moeller
Hi Joel, > On Dec 17, 2021, at 09:36, Joel Wirāmu Pauling wrote: > > The XG PON ONT units from Nokia/Huawei are coming with only 10G NbaseT > (usually singular port) only in the consumer access space. No SFP+ Yes, ISPs are not that keen on distributing SFP/SFP+ modules, but there are some C

Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Sebastian Moeller
Mmmh, I guess our approach at traffic shaping does not scale well at those speeds. Maybe this could be fixed with larger batching? I think it might be worth trying to switch to simple.qos/fq_codel and set a somewhat larger burst/quantum duration in defaults.sh, then disable BQL on the NIC and c

Re: [Cerowrt-devel] 10gige and 2.5gige

2021-12-17 Thread Toke Høiland-Jørgensen via Cerowrt-devel
--- Begin Message --- Sebastian Moeller writes: > Mmmh, I guess our approach at traffic shaping does not scale well at those > speeds. Maybe this could be fixed with larger batching? > > I think it might be worth trying to switch to simple.qos/fq_codel and > set a somewhat larger burst/quantum d