Thread (29 messages) 29 messages, 9 authors, 2014-09-01

Re: [PATCH 0/2] Get rid of ndo_xmit_flush

From: Jesper Dangaard Brouer <hidden>
Date: 2014-08-27 08:37:44

On Wed, 27 Aug 2014 09:48:59 +0200
Jesper Dangaard Brouer [off-list ref] wrote:
Reading Jeremy Eder's blog post:
 http://www.breakage.org/2012/11/14/processor-max_cstate-intel_idle-max_cstate-and-devcpu_dma_latency/

It seems the best method for disabling these C and P states, and
keeping all CPUs in C0/C1 state is doing:

# tuned-adm profile latency-performance
[...]
I'm also going to disable Hyper-threading in BIOS, even-though I'm only
using one CPU in these tests (I worry that some process running on a
sibling could disturb the accuracy).
Disabling Hyper-Threading in BIOS helped the accuracy, now the latency
variation is below 0.1ns min:-0.031 and max:0.054 over 120 sec.

Still using:
 # tuned-adm profile latency-performance
Setting:
 * sudo ethtool -C eth5 rx-usecs 30
Result:
 * instant rx:1 tx:5603644 pps n:120 average: rx:0 tx:5603317 pps
   (instant variation TX 0.010 ns (min:-0.132 max:0.114) RX 0.000 ns)
 * instant rx:1 tx:5599820 pps n:300 average: rx:1 tx:5602382 pps
   (instant variation TX -0.082 ns (min:-0.209 max:0.114) RX 0.000 ns)
With no HT:
 * instant rx:1 tx:5597616 pps n:120 average: rx:1 tx:5596283 pps
  (instant variation TX 0.043 ns (min:-0.031 max:0.054) RX 0.000 ns)

With this setup, I think we are ready to measure the effect of our API
changes with sufficient accuracy.

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Sr. Network Kernel Developer at Red Hat
  Author of http://www.iptv-analyzer.org
  LinkedIn: http://www.linkedin.com/in/brouer
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help