Raw dump of memcached benchmark results

Initial benchmarks were run on an Arndale, expected to run also on Calxeda.

Server and client running on the same machine

Server exposing a 256 MB cache over a UNIX socket

root@arndale:~# memcached -u memcache -s /tmp/memcache -m 256

Single thread

Client:

root@arndale:~# memcslap --servers=/tmp/memcache --concurrency=1 --execute-number=100000 --binary
        Threads connecting to servers 1
        Took 43.891 seconds to load data

Trace:

+  28.91%  memcached  [kernel.kallsyms]      [k] _raw_spin_unlock_irqrestore
+   7.93%  memcached  [kernel.kallsyms]      [k] lock_acquire
+   3.24%  memcached  [kernel.kallsyms]      [k] _raw_spin_unlock_irq
+   3.05%  memcached  [kernel.kallsyms]      [k] __mutex_unlock_slowpath
+   2.88%  memcached  [kernel.kallsyms]      [k] mutex_lock_nested
+   2.62%  memcached  [kernel.kallsyms]      [k] lock_release
+   2.56%  memcached  [kernel.kallsyms]      [k] vector_swi
+   2.17%  memcached  memcached              [.] event_handler
+   1.75%  memcached  libpthread-2.15.so     [.] __pthread_mutex_unlock_usercnt
+   1.41%  memcached  [kernel.kallsyms]      [k] fget_light
+   1.32%  memcached  [kernel.kallsyms]      [k] kfree
+   1.29%  memcached  [kernel.kallsyms]      [k] mutex_lock_interruptible_nested
+   1.27%  memcached  libc-2.15.so           [.] vfprintf
+   1.20%  memcached  libpthread-2.15.so     [.] pthread_mutex_trylock
+   1.08%  memcached  memcached              [.] hash
+   0.97%  memcached  libc-2.15.so           [.] memcpy
+   0.87%  memcached  libpthread-2.15.so     [.] pthread_mutex_lock
    0.86%  memcached  libevent-2.0.so.5.1.7  [.] event_base_loop

Multiple threads

Client:

root@arndale:~# memcslap --servers=/tmp/memcache --concurrency=16 --execute-number=100000 --binary
        Threads connecting to servers 16
        Took 117.046 seconds to load data

Trace:

+  34.86%  memcached  libpthread-2.15.so     [.] pthread_mutex_trylock
+  17.92%  memcached  [kernel.kallsyms]      [k] _raw_spin_unlock_irqrestore
+   5.68%  memcached  [kernel.kallsyms]      [k] lock_acquire
+   1.91%  memcached  [kernel.kallsyms]      [k] lock_release
+   1.89%  memcached  [kernel.kallsyms]      [k] _raw_spin_unlock_irq
+   1.55%  memcached  memcached              [.] event_handler
+   1.54%  memcached  libpthread-2.15.so     [.] __pthread_mutex_unlock_usercnt
+   1.37%  memcached  [kernel.kallsyms]      [k] vector_swi
+   1.19%  memcached  memcached              [.] do_item_alloc
+   1.15%  memcached  [kernel.kallsyms]      [k] mutex_lock_interruptible_nested
+   1.14%  memcached  [kernel.kallsyms]      [k] __mutex_unlock_slowpath
+   1.05%  memcached  libpthread-2.15.so     [.] pthread_mutex_lock
+   0.99%  memcached  [kernel.kallsyms]      [k] kfree
+   0.95%  memcached  memcached              [.] hash
+   0.91%  memcached  libc-2.15.so           [.] vfprintf
+   0.81%  memcached  [kernel.kallsyms]      [k] kmem_cache_free
+   0.75%  memcached  [kernel.kallsyms]      [k] fget_light
+   0.72%  memcached  [kernel.kallsyms]      [k] exynos4_frc_read
+   0.68%  memcached  memcached              [.] assoc_find
+   0.59%  memcached  [kernel.kallsyms]      [k] unix_stream_recvmsg
+   0.58%  memcached  libc-2.15.so           [.] memcpy
+   0.57%  memcached  [kernel.kallsyms]      [k] __kmalloc_track_caller

LEG/Engineering/vertical-web/memcached-manual-benchmarks (last modified 2013-04-03 19:20:18)