Jitendra Kolhe 1e356fc14b mem-prealloc: reduce large guest start-up and migration time. ... Using "-mem-prealloc" option for a large guest leads to higher guest start-up and migration time. This is because with "-mem-prealloc" option qemu tries to map every guest page (create address translations), and make sure the pages are available during runtime. virsh/libvirt by default, seems to use "-mem-prealloc" option in case the guest is configured to use huge pages. The patch tries to map all guest pages simultaneously by spawning multiple threads. Currently limiting the change to QEMU library functions on POSIX compliant host only, as we are not sure if the problem exists on win32. Below are some stats with "-mem-prealloc" option for guest configured to use huge pages. ------------------------------------------------------------------------ Idle Guest | Start-up time | Migration time ------------------------------------------------------------------------ Guest stats with 2M HugePage usage - single threaded (existing code) ------------------------------------------------------------------------ 64 Core - 4TB | 54m11.796s | 75m43.843s 64 Core - 1TB | 8m56.576s | 14m29.049s 64 Core - 256GB | 2m11.245s | 3m26.598s ------------------------------------------------------------------------ Guest stats with 2M HugePage usage - map guest pages using 8 threads ------------------------------------------------------------------------ 64 Core - 4TB | 5m1.027s | 34m10.565s 64 Core - 1TB | 1m10.366s | 8m28.188s 64 Core - 256GB | 0m19.040s | 2m10.148s ----------------------------------------------------------------------- Guest stats with 2M HugePage usage - map guest pages using 16 threads ----------------------------------------------------------------------- 64 Core - 4TB | 1m58.970s | 31m43.400s 64 Core - 1TB | 0m39.885s | 7m55.289s 64 Core - 256GB | 0m11.960s | 2m0.135s ----------------------------------------------------------------------- Changed in v2: - modify number of memset threads spawned to min(smp_cpus, 16). - removed 64GB memory restriction for spawning memset threads. Changed in v3: - limit number of threads spawned based on min(sysconf(_SC_NPROCESSORS_ONLN), 16, smp_cpus) - implement memset thread specific siglongjmp in SIGBUS signal_handler. Changed in v4 - remove sigsetjmp/siglongjmp and SIGBUS unblock/block for main thread as main thread no longer touches any pages. - simplify code my returning memset_thread_failed status from touch_all_pages. Signed-off-by: Jitendra Kolhe <jitendra.kolhe@hpe.com> Message-Id: <1487907103-32350-1-git-send-email-jitendra.kolhe@hpe.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>		2017-03-14 13:26:36 +01:00
..
acl.c
aio-posix.c	async: remove unnecessary inc/dec pairs	2017-02-21 11:39:40 +00:00
aio-win32.c	async: remove unnecessary inc/dec pairs	2017-02-21 11:39:40 +00:00
aiocb.c	block: move AioContext, QEMUTimer, main-loop to libqemuutil	2017-02-21 11:14:07 +00:00
async.c	async: remove unnecessary inc/dec pairs	2017-02-21 11:39:40 +00:00
base64.c
bitmap.c	bitmap: assert that start and nr are non negative	2017-01-20 13:22:17 +01:00
bitops.c
buffer.c
bufferiszero.c
compatfd.c	cpus: remove ugly cast on sigbus_handler	2017-03-03 16:40:02 +01:00
coroutine-gthread.c
coroutine-sigaltstack.c
coroutine-ucontext.c
coroutine-win32.c
crc32c.c
cutils.c	util/cutils: Change qemu_strtosz*() from int64_t to uint64_t	2017-02-23 20:35:36 +01:00
envlist.c
error.c
event_notifier-posix.c	event_notifier: cleanups around event_notifier_set_handler	2017-01-16 17:52:35 +01:00
event_notifier-win32.c	event_notifier: cleanups around event_notifier_set_handler	2017-01-16 17:52:35 +01:00
fifo8.c	migration: consolidate VMStateField.start	2017-02-13 17:27:13 +00:00
getauxval.c
hbitmap.c	hbitmap: Add hbitmap_is_serializable()	2017-01-26 10:25:01 +08:00
hexdump.c
host-utils.c	host-utils: Implement unsigned quadword left/right shift and unit tests	2017-01-31 10:10:14 +11:00
id.c
iohandler.c	block: move AioContext, QEMUTimer, main-loop to libqemuutil	2017-02-21 11:14:07 +00:00
iov.c
keyval.c	keyval: Support lists	2017-03-07 16:07:48 +01:00
lockcnt.c	qemu-thread: optimize QemuLockCnt with futexes on Linux	2017-01-16 13:25:18 +00:00
log.c	util/cutils: Rename qemu_strtoll(), qemu_strtoull()	2017-02-23 20:35:35 +01:00
main-loop.c	cpus: remove ugly cast on sigbus_handler	2017-03-03 16:40:02 +01:00
Makefile.objs	keyval: New keyval_parse()	2017-03-07 16:07:46 +01:00
memfd.c
mmap-alloc.c	exec, kvm, target-ppc: Move getrampagesize() to common code	2017-03-03 11:30:59 +11:00
module.c
notify.c
osdep.c
oslib-posix.c	mem-prealloc: reduce large guest start-up and migration time.	2017-03-14 13:26:36 +01:00
oslib-win32.c	mem-prealloc: reduce large guest start-up and migration time.	2017-03-14 13:26:36 +01:00
path.c
qdist.c
qemu-config.c
qemu-coroutine-io.c
qemu-coroutine-lock.c	coroutine-lock: make CoRwlock thread-safe and fair	2017-02-21 11:39:40 +00:00
qemu-coroutine-sleep.c	block: explicitly acquire aiocontext in timers that need it	2017-02-21 11:14:08 +00:00
qemu-coroutine.c	coroutine-lock: add limited spinning to CoMutex	2017-02-21 11:39:40 +00:00
qemu-error.c	qemu-error: remove dependency of stubs on monitor	2016-11-01 16:06:57 +01:00
qemu-openpty.c
qemu-option.c	option: Tweak invalid size error message and unbreak iotest 049	2017-02-28 20:40:31 +01:00
qemu-progress.c
qemu-sockets.c	io: introduce a DNS resolver API	2017-01-23 15:32:46 +00:00
qemu-thread-posix.c	* QOM interface fix (Eduardo)	2017-01-20 16:42:07 +00:00
qemu-thread-win32.c	* QOM interface fix (Eduardo)	2017-01-20 16:42:07 +00:00
qemu-timer-common.c
qemu-timer.c	timer: use an inline function for free	2017-03-01 00:09:28 +04:00
qht.c
range.c
rcu.c
readline.c
thread-pool.c	block: explicitly acquire aiocontext in aio callbacks that need it	2017-02-21 11:39:39 +00:00
throttle.c
timed-average.c
trace-events	coroutine-lock: make CoMutex thread-safe	2017-02-21 11:39:40 +00:00
unicode.c
uri.c	Fix documentation and some comments (article, grammar)	2017-01-24 23:26:52 +03:00
uuid.c