sysctl: Adopt more settings from the IBM HPC guidelines

https://www.ibm.com/developerworks/community/wikis/home?lang=en#!/wiki/Welcome%20to%20High%20Performance%20Computing%20%28HPC%29%20Central/page/Linux%20System%20Tuning%20Recommendations

Since we have already configured most of our IP/TCP stack
for low latency and fast throughput, these settings complete
those efforts.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>
Signed-off-by: Arne Fitzenreiter <arne_f@ipfire.org>

diff --git a/config/etc/sysctl.conf b/config/etc/sysctl.conf
index 9a943fffa886430d4d45078f19f5b3dbc796bbb6..d11e53c88db373ac410f4490a492d4cf37d737d8 100644 (file)
--- a/config/etc/sysctl.conf
+++ b/config/etc/sysctl.conf
@@ -76,9 +76,16 @@ vm.swappiness = 1
 kernel.sched_migration_cost_ns = 5000000
 
 # Increase kernel buffer size maximums
+net.ipv4.tcp_mem = 16777216 16777216 16777216
 net.ipv4.tcp_rmem = 4096 87380 16777216
 net.ipv4.tcp_wmem = 4096 16384 16777216
 net.ipv4.udp_mem = 3145728 4194304 16777216
 
+# Prefer low latency over higher throughput
+net.ipv4.tcp_low_latency=1
+
+# Reserve more socket space for the TCP window
+net.ipv4.tcp_adv_win_scale=2
+
 # Enable TCP fast-open
 net.ipv4.tcp_fastopen = 3