How are you measuring that?IPerf plug-in on FW or IPerf.server-DEC3840-IPerf.client?
* pppoe wan* 10gIn my experience that did not work well with high bandwith, no matter if FreeBSD or Linux, pppoe has way to much overhead for that.
I see...Directly plugged into DEC3840 or via Switch (if so, can you name make, model and SW version pls)?
Quote from: Cerberus on May 24, 2022, 02:56:45 pm* pppoe wan* 10gIn my experience that did not work well with high bandwith, no matter if FreeBSD or Linux, pppoe has way to much overhead for that.I don't disagree with that, however, as part of my testing, I've had my ISP to provide me with DHCP instead of PPPoE and guess what? Situation remains unchanged...
Can you what tuning have you applied to increase performances?
Turn on VLAN Hardware Offload, enable the parent interface (no IPs or anything, just assign it and enable), also enabled HT interrupts as per https://wiki.freebsd.org/Networking/10GbE/Router#Allow_interrupts_on_HTT_logical_CPUs this BSD 10Gb tuning article
Quote from: Berzerker on May 25, 2022, 04:49:36 amTurn on VLAN Hardware Offload, enable the parent interface (no IPs or anything, just assign it and enable), also enabled HT interrupts as per https://wiki.freebsd.org/Networking/10GbE/Router#Allow_interrupts_on_HTT_logical_CPUs this BSD 10Gb tuning articleI'm not sure how to enable HTT interrupts, however, would it make any sense at all since DEC3840 is AMD based whilst HT instructions are Intel's?
I don't think there are significant changes on 22.1.x and I'm not sure why people sometimes wonder if there are... release notes state what is being worked on and source repositories have annotated changes too.As a general principle: performance gain is nice, but stability is much nicer still.Cheers,Franco
hw.ibrs_disable:1net.inet.rss.enabled:1net.isr.maxthreads:-1
All fine and well, but we're still (I am at least, I presume others are too) experiencing some performance issues.
Quote from: Berzerker on May 25, 2022, 08:48:42 pmAll fine and well, but we're still (I am at least, I presume others are too) experiencing some performance issues.Forgive me for missing the full context here. I can't judge your setup from here, but I would assume the performance numbers given are rooted in reality for both the specifications and your measurements. The bigger question is who is going to verify why these values differ and what could be done about it.Cheers,Franco