diff options
author | Jesper Dangaard Brouer <brouer@redhat.com> | 2015-05-21 13:18:12 +0300 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2015-05-23 06:59:17 +0300 |
commit | 1d73ba16ad3975a2eb27e47e52fc215c0c5b8d83 (patch) | |
tree | e13418ede08a6f78594389094a1d66ea17469887 /samples/pktgen | |
parent | 282fb58947e129dea8badf48972ea89d027a76dc (diff) | |
download | linux-1d73ba16ad3975a2eb27e47e52fc215c0c5b8d83.tar.xz |
pktgen: add sample script pktgen_sample03_burst_single_flow.sh
Add the pktgen samples script pktgen_sample03_burst_single_flow.sh
that demonstrates how to acheive maximum performance.
If correctly tuned[1] single CPU 10Gbit/s wirespeed small pkts is
possible[2] which is 14.88Mpps. The trick is to take advantage of the
"burst" feature introduced in commit 38b2cf2982dc73 ("net: pktgen:
packet bursting via skb->xmit_more").
[1] http://netoptimizer.blogspot.dk/2014/06/pktgen-for-network-overload-testing.html
[2] http://netoptimizer.blogspot.dk/2014/10/unlocked-10gbps-tx-wirespeed-smallest.html
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'samples/pktgen')
-rwxr-xr-x | samples/pktgen/pktgen_sample03_burst_single_flow.sh | 82 |
1 files changed, 82 insertions, 0 deletions
diff --git a/samples/pktgen/pktgen_sample03_burst_single_flow.sh b/samples/pktgen/pktgen_sample03_burst_single_flow.sh new file mode 100755 index 000000000000..775f5d0a1e53 --- /dev/null +++ b/samples/pktgen/pktgen_sample03_burst_single_flow.sh @@ -0,0 +1,82 @@ +#!/bin/bash +# +# Script for max single flow performance +# - If correctly tuned[1], single CPU 10G wirespeed small pkts is possible[2] +# +# Using pktgen "burst" option (use -b $N) +# - To boost max performance +# - Avail since: kernel v3.18 +# * commit 38b2cf2982dc73 ("net: pktgen: packet bursting via skb->xmit_more") +# - This avoids writing the HW tailptr on every driver xmit +# - The performance boost is impressive, see commit and blog [2] +# +# Notice: On purpose generates a single (UDP) flow towards target, +# reason behind this is to only overload/activate a single CPU on +# target host. And no randomness for pktgen also makes it faster. +# +# Tuning see: +# [1] http://netoptimizer.blogspot.dk/2014/06/pktgen-for-network-overload-testing.html +# [2] http://netoptimizer.blogspot.dk/2014/10/unlocked-10gbps-tx-wirespeed-smallest.html +# +basedir=`dirname $0` +source ${basedir}/functions.sh +root_check_run_with_sudo "$@" + +# Parameter parsing via include +source ${basedir}/parameters.sh +# Set some default params, if they didn't get set +[ -z "$DEST_IP" ] && DEST_IP="198.18.0.42" +[ -z "$DST_MAC" ] && DST_MAC="90:e2:ba:ff:ff:ff" +[ -z "$BURST" ] && BURST=32 +[ -z "$CLONE_SKB" ] && CLONE_SKB="100000" + +# Base Config +DELAY="0" # Zero means max speed +COUNT="0" # Zero means indefinitely + +# General cleanup everything since last run +pg_ctrl "reset" + +# Threads are specified with parameter -t value in $THREADS +for ((thread = 0; thread < $THREADS; thread++)); do + dev=${DEV}@${thread} + + # Add remove all other devices and add_device $dev to thread + pg_thread $thread "rem_device_all" + pg_thread $thread "add_device" $dev + + # Base config + pg_set $dev "flag QUEUE_MAP_CPU" + pg_set $dev "count $COUNT" + pg_set $dev "clone_skb $CLONE_SKB" + pg_set $dev "pkt_size $PKT_SIZE" + pg_set $dev "delay $DELAY" + pg_set $dev "flag NO_TIMESTAMP" + + # Destination + pg_set $dev "dst_mac $DST_MAC" + pg_set $dev "dst $DEST_IP" + + # Setup burst, for easy testing -b 0 disable bursting + # (internally in pktgen default and minimum burst=1) + if [[ ${BURST} -ne 0 ]]; then + pg_set $dev "burst $BURST" + else + info "$dev: Not using burst" + fi +done + +# Run if user hits control-c +function control_c() { + # Print results + for ((thread = 0; thread < $THREADS; thread++)); do + dev=${DEV}@${thread} + echo "Device: $dev" + cat /proc/net/pktgen/$dev | grep -A2 "Result:" + done +} +# trap keyboard interrupt (Ctrl-C) +trap control_c SIGINT + +echo "Running... ctrl^C to stop" >&2 +pg_ctrl "start" |