mdcached - A high-performance cache daemon (Bullet Cache)
*mdcached* -h
*mdcached* [-h] [-a] [-d] [-f file] [-i] [-k] [-m #] [-n #] [-r] [-s #] [-t #] [-v #]
The optional argument to mdcached are:
-a
Disable runtime monitoring and adaptive reconfiguration
of the server.
-d
Enable debug mode: the server does not fork to background
and displays the log messages on the terminal.
-f
file Set dump file name (for use with prewarming or with
SIGUSR1). The server has the ability to create a dump file containing stored
data. This dump may be created by sending the SIGUSR1 signal to the server,
and it may be used to prewarm the server’s cache.
-h
Shows a help message with a short summary of available
command line arguments
-i #
Automatically create a cache dump every # seconds.
-k
Bind network threads to worker threads. This option
enables a performance optimization which may increase performance when worker
threads are enabled. The most efficient case is when the number of worker
threads equals the number of network threads.
-m #
Limits memory use to # MiB (default=0, disabled).
Enabling memory limits may hurt performance in some cases. The limits are only
enforced for the raw data stored in the memory database, and the actual memory
usage (which includes internal book-keeping data) may be significantly
higher
-n #
Set number of network threads to # (default=2). If
Bullet Cache is used on dedicated servers, the most efficient
configuration is to configure the number of network threads between 1 and 1.5
times the number of logical CPUs in the system.
-r
Read cache data from the dump file (prewarm the
cache).
-s #
Set the default record size, in bytes. Correctly
specifying the record size may result in significantly lower memory
usage.
-t #
Set number of worker threads to # (default=0). Worker
threads are useful only if the number of complex data queries is larger than
the number of simple data queries. For normal workloads, worker threads are
likely to reduce the overall performance.
-v #
Sets minimum log (verbosity) level to # [0-7]. The log
level semantics (and numbering) resembles the ones used in
syslog(3).
Bullet Cache is a memory database intended to be used much
like memcached, but offering much higher flexibility through use of record
tags which can be used to perform bulk operations on sets of records. It has
its own binary network protocol which is accessible through Unix domain
sockets and through TCP. It is indended to be accessed through a
high-performance C library.
This documentation is yet to be completed.
Bullet Cache has extensive documentation available at:
•
http://mdcached.sourceforge.net/documentation.html
• http://goo.gl/LTrzt
Please report bugs and satisfied user reports at the SourceForge
project page (https://sourceforge.net/projects/mdcached/).
Bullet Cache and its documentation was created by Ivan Voras
<ivoras@gmail.com>.
Visit the GSP FreeBSD Man Page Interface.