[ntp:questions] Interesting results with PPS

Rick Jones rick.jones2 at hp.com
Wed Feb 7 18:45:08 UTC 2007


I am _definitely_ from the back of the peanut gallery when it comes to
timekeeping, but since someone has mentioned networks, and since I
have a little experience with that, at the risk of restating something
already known to the timekeepers I'll point-out that many NIC's today
have interrupt coalescing parameters, which are often set by default
to favor CPU utilization and maximum throughput over latency.  Some of
these mechanisms are based on timers running in the NIC.  Their effect
can be seen in things like netperf TCP_RR tests.

ftp://ftp.cup.hp.com/dist/networking/briefs/nic_latency_vs_tput.txt

has some jumbled results of what can change based on some tests I've
run in the past.  

hth,

rick jones
-- 
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...




More information about the questions mailing list