[RFC] TUNING: document THP caveat for Linux users

[RFC] TUNING: document THP caveat for Linux users

[Eric Wong]

This probably applies to other kernels, too, but I'm most
familiar with Linux.
---
 It took me a while to get the wording below to this point.
 Maybe there's not enough detail for folks unfamiliar with
 how OSes work, or maybe there's too much and will be TL;DR-ed...

 TUNING | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

diff --git a/TUNING b/TUNING
index 247090b..1f55228 100644
--- a/TUNING
+++ b/TUNING
@@ -72,10 +72,28 @@ See Unicorn::Configurator for details on the config file format.
   have them unbuffered (File#sync = true) or they are
   record(line)-buffered in userspace before any writes.
 
-== Kernel Parameters (Linux sysctl)
+== Kernel Parameters (Linux sysctl and sysfs)
 
 WARNING: Do not change system parameters unless you know what you're doing!
 
+* Transparent hugepages (THP) improves performance in many cases,
+  but can also increase memory use when relying on a
+  copy-on-write(CoW)-friendly GC (Ruby 2.0+) with "preload_app true".
+  CoW operates at the page level, so writing to a huge page would
+  trigger a 2 MB copy (x86-64), as opposed to a 4 KB copy on a
+  regular (non-huge) page.
+
+  Consider only allowing THP to be used when requested via the
+  madvise(2) syscall:
+
+	echo madvise >/sys/kernel/mm/transparent_hugepage/enabled
+
+  Or disabling it system-wide, via "never".
+
+  n.b. "page" in this context only applies to the OS kernel,
+  Ruby GC implementations also use this term for the same concept
+  in a way that is agnostic to the OS.
+
 * net.core.rmem_max and net.core.wmem_max can increase the allowed
   size of :rcvbuf and :sndbuf respectively. This is mostly only useful
   for UNIX domain sockets which do not have auto-tuning buffer sizes.
-- 
EW