<div dir="ltr"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">This comes from OVS code and shows OVS thread spinning, not DPDK PMD.<br>Blame the OVS application for not using e.g. _mm_pause() and burning<br>the CPU like crazy.<br></blockquote><div><br></div><div><br></div>OK, I'm citing a bit more from the same reference:<br><div><b>"By tracing back to the function’s caller </b></div><div><b>in the PMD thread main(void *f_), </b></div><div>we found that the thread kept spinning on the following code block:</div><div><div>for ( ; ; ) {<br>for ( i = 0; i < poll_cnt; i ++) {<br>dp_netdev_process_rxq_port (pmd, list[i].port, poll_list[i].rx) ;<br>}<br>}</div>This indicates that the [PMD] thread was continuously<br>monitoring and executing the receiving data path." </div><div><br></div><div>Cheers,</div><div><br></div><div>Etienne</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Feb 23, 2021 at 10:33 PM Pawel Malachowski <<a href="mailto:pawmal-nanog@freebsd.lublin.pl">pawmal-nanog@freebsd.lublin.pl</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">> > No, it is not PMD that runs the processor in a polling loop.<br>
> > It is the application itself, thay may or may not busy loop,<br>
> > depending on application programmers choice.<br>
> <br>
> From one of my earlier references [2]:<br>
> <br>
> "we found that a poll mode driver (PMD)<br>
> thread accounted for approximately 99.7 percent<br>
> CPU occupancy (a full core utilization)."<br>
> <br>
> And further on:<br>
> <br>
> "we found that the thread kept spinning on the following code block:<br>
> <br>
> *for ( ; ; ) {for ( i = 0; i < poll_cnt; i ++) {dp_netdev_process_rxq_port<br>
> (pmd, list[i].port, poll_list[i].rx) ;}}*<br>
> This indicates that the thread was continuously<br>
> monitoring and executing the receiving data path."<br>
<br>
This comes from OVS code and shows OVS thread spinning, not DPDK PMD.<br>
Blame the OVS application for not using e.g. _mm_pause() and burning<br>
the CPU like crazy.<br>
<br>
<br>
For comparison, take a look at top+i7z output from DPDK-based 100G DDoS<br>
scrubber currently lifting some low traffic using cores 1-13 on 16 core<br>
host. It uses naive DPDK::rte_pause() throttling to enter C1.<br>
<br>
Tasks: 342 total, 1 running, 195 sleeping, 0 stopped, 0 zombie<br>
%Cpu(s): 6.6 us, 0.6 sy, 0.0 ni, 89.7 id, 3.1 wa, 0.0 hi, 0.0 si, 0.0 st<br>
<br>
Core [core-id] :Actual Freq (Mult.) C0% Halt(C1)% C3 % C6 % Temp VCore<br>
Core 1 [0]: 1467.73 (14.68x) 2.15 5.35 1 92.3 43 0.6724<br>
Core 2 [1]: 1201.09 (12.01x) 11.7 93.9 0 0 39 0.6575<br>
Core 3 [2]: 1200.06 (12.00x) 11.8 93.8 0 0 42 0.6543<br>
Core 4 [3]: 1200.14 (12.00x) 11.8 93.8 0 0 41 0.6549<br>
Core 5 [4]: 1200.10 (12.00x) 11.8 93.8 0 0 41 0.6526<br>
Core 6 [5]: 1200.12 (12.00x) 11.8 93.8 0 0 40 0.6559<br>
Core 7 [6]: 1201.01 (12.01x) 11.8 93.8 0 0 41 0.6559<br>
Core 8 [7]: 1201.02 (12.01x) 11.8 93.8 0 0 43 0.6525<br>
Core 9 [8]: 1201.00 (12.01x) 11.8 93.8 0 0 41 0.6857<br>
Core 10 [9]: 1201.04 (12.01x) 11.8 93.8 0 0 40 0.6541<br>
Core 11 [10]: 1201.95 (12.02x) 13.6 92.9 0 0 40 0.6558<br>
Core 12 [11]: 1201.02 (12.01x) 11.8 93.8 0 0 42 0.6526<br>
Core 13 [12]: 1204.97 (12.05x) 17.6 90.8 0 0 45 0.6814<br>
Core 14 [13]: 1248.39 (12.48x) 28.2 84.7 0 0 41 0.6855<br>
Core 15 [14]: 2790.74 (27.91x) 91.9 0 1 1 41 0.8885 <-- not PMD<br>
Core 16 [15]: 1262.29 (12.62x) 13.1 34.9 1.7 56.2 43 0.6616<br>
<br>
$ dataplanectl stats fcore | grep total<br>
fcore total idle 393788223887 work 860443658 (0.2%) (forced-idle 7458486526622) recv 202201388561 drop 61259353721 (30.3%) limit 269909758 (0.1%) pass 140606076622 (69.6%) ingress 66048460 (0.0%/0.0%) sent 162580376914 (80.4%/100.0%) overflow 0 (0.0%) sampled 628488188/628488188<br>
<br>
<br>
<br>
-- <br>
Pawel Malachowski<br>
@pawmal80<br>
</blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr"><span style="color:rgb(136,136,136);font-family:tahoma,sans-serif">Ing. Etienne-Victor Depasquale</span><br style="color:rgb(136,136,136);font-family:tahoma,sans-serif"><span style="color:rgb(136,136,136);font-family:tahoma,sans-serif">Assistant Lecturer</span><br style="color:rgb(136,136,136);font-family:tahoma,sans-serif"><span style="color:rgb(136,136,136);font-family:tahoma,sans-serif">Department of Communications & Computer Engineering</span><br style="color:rgb(136,136,136);font-family:tahoma,sans-serif"><span style="color:rgb(136,136,136);font-family:tahoma,sans-serif">Faculty of Information & Communication Technology</span><br style="color:rgb(136,136,136);font-family:tahoma,sans-serif"><span style="color:rgb(136,136,136);font-family:tahoma,sans-serif">University of Malta</span><div>Web. <a href="https://www.um.edu.mt/profile/etiennedepasquale" target="_blank">https://www.um.edu.mt/profile/etiennedepasquale</a><br></div></div></div>