<div dir="ltr">Hi Dan,<div>We obtained great numbers with only one instance after the changes that Remi mentioned in 1.3.X</div><div><br></div><div>Almost 600Kqps as mentioned.</div><div>The setup involved disabling all FW rules, some sysctls and ethtool, but no cpu affinity.</div><div><br></div><div>Don't have all the setup here now, But its also very specific to your setup.</div><div><br></div><div>Good luck!</div><div>nico</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Jan 3, 2023 at 1:19 PM Remi Gacogne via dnsdist <<a href="mailto:dnsdist@mailman.powerdns.com">dnsdist@mailman.powerdns.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 03/01/2023 17:07, Dan McCombs wrote:<br>
> I spent some time looking through the thread Nicolas shared at <br>
> <a href="https://mailman.powerdns.com/pipermail/dnsdist/2017-April/000281.html" rel="noreferrer" target="_blank">https://mailman.powerdns.com/pipermail/dnsdist/2017-April/000281.html</a> <br>
> <<a href="https://mailman.powerdns.com/pipermail/dnsdist/2017-April/000281.html" rel="noreferrer" target="_blank">https://mailman.powerdns.com/pipermail/dnsdist/2017-April/000281.html</a>> <br>
> of his performance tuning experience. Do you know if it's still the case <br>
> in newer versions of dnsdist to need multiple instances of dnsdist <br>
> running on a machine this size (32 cores) rather than just additional <br>
> listeners via addLocal to really push past 400kQPS? I just want to make <br>
> sure I'm not working off out-of-date information.<br>
<br>
A lot has changed with regard to scalability on machines with a great <br>
number of cores. I remember us doing a lot of work in that area in <br>
1.3.0, and we also added a few more improvements since then (false <br>
sharing work-around, more sharding, more efficient locking, etc)<br>
<br>
I have personally reached quite more than 400k QPS with a single dnsdist <br>
process during my tests, but you might be able to get better performance <br>
with several processes as we are still sharing some state between <br>
threads. For example Otto has been working on more scalable counters in <br>
the recursor lately, and this is something that I would like to <br>
implement in dnsdist. It will likely have to wait until 1.9, so around <br>
Q3 this year, unless someone beats me to it :)<br>
<br>
Please report back with your findings, as it is really helpful to hear <br>
about different setups than the ones we already do our performance <br>
testing on.<br>
<br>
-- <br>
Remi Gacogne<br>
PowerDNS.COM BV - <a href="https://www.powerdns.com/" rel="noreferrer" target="_blank">https://www.powerdns.com/</a><br>
<br>
_______________________________________________<br>
dnsdist mailing list<br>
<a href="mailto:dnsdist@mailman.powerdns.com" target="_blank">dnsdist@mailman.powerdns.com</a><br>
<a href="https://mailman.powerdns.com/mailman/listinfo/dnsdist" rel="noreferrer" target="_blank">https://mailman.powerdns.com/mailman/listinfo/dnsdist</a><br>
</blockquote></div>