Initial commit of HPC compute profile

profiles/hpc-compute/tuned.conf

@@ -0,0 +1,37 @@
+#
+# tuned configuration
+#
+
+[main]
+summary=Optimize for HPC compute workloads
+include=throughput-performance
+
+[sysctl]
+# Busy polling helps reduce latency in the network receive path
+# by allowing socket layer code to poll the receive queue of a
+# network device, and disabling network interrupts.
+# busy_read value greater than 0 enables busy polling. Recommended
+# net.core.busy_read value is 50.
+# busy_poll value greater than 0 enables polling globally. 
+# Recommended net.core.busy_poll value is 50 
+net.core.busy_read=50
+net.core.busy_poll=50
+
+# TCP fast open reduces network latency by enabling data exchange
+# during the sender's initial TCP SYN. The value 3 enables fast open
+# on client and server connections.
+net.ipv4.tcp_fastopen=3
+
+# Most HPC applications have NUMA knowledge. Disabling automatic NUMA
+# balancing prevents unwanted memory unmapping. 
+kernel.numa_balancing=0
+
+# The total time the scheduler will consider a migrated process
+# "cache hot" and thus less likely to be re-migrated
+# (system default is 500000, i.e. 0.5 ms)
+kernel.sched_migration_cost_ns=5000000
+
+[vm]
+# Most HPC applications can take advantage of large page sizes.
+# This setting ensures that THP is forced on.
+transparent_hugepages=always