SharedCache(3) User Contributed Perl Documentation SharedCache(3)NAMEIPC::SharedCache - a Perl module to manage a cache in SysV IPC shared
memory.
SYNOPSIS
use IPC::SharedCache;
# the cache is accessed using a tied hash.
tie %cache, 'IPC::SharedCache', ipc_key => 'AKEY',
load_callback => \&load,
validate_callback => \&validate;
# get an item from the cache
$config_file = $cache{'/some/path/to/some.config'};
DESCRIPTION
This module provides a shared memory cache accessed as a tied hash.
Shared memory is an area of memory that is available to all processes.
It is accessed by choosing a key, the ipc_key arguement to tie. Every
process that accesses shared memory with the same key gets access to
the same region of memory. In some ways it resembles a file system,
but it is not hierarchical and it is resident in memory. This makes it
harder to use than a filesystem but much faster. The data in shared
memory persists until the machine is rebooted or it is explicitely
deleted.
This module attempts to make shared memory easy to use for one specific
application - a shared memory cache. For other uses of shared memory
see the documentation to the excelent module I use, IPC::ShareLite
(IPC::ShareLite).
A cache is a place where processes can store the results of their
computations for use at a later time, possibly by other instances of
the application. A good example of the use of a cache is a web server.
When a web server receieves a request for an html page it goes to the
file system to read it. This is pretty slow, so the web server will
probably save the file in memory and use the in memory copy the next
time a request for that file comes in, as long as the file hasn't
changed on disk. This certainly speeds things up but web servers have
to serve multiple clients at once, and that means multiple copies of
the in-memory data. If the web server uses a shared memory cache, like
the one this module provides, then all the servers can use the same
cache and much less memory is consumed.
This module handles all shared memory interaction using the
IPC::ShareLite module (version 0.06 and higher) and all data
serialization using Storable. See IPC::ShareLite and Storable for
details.
MOTIVATION
This module began its life as an internal piece of HTML::Template (see
HTML::Template). HTML::Template has the ability to maintain a cache of
parsed template structures when running in a persistent environment
like Apache/mod_perl. Since parsing a template from disk takes a fair
ammount of time this can provide a big performance gain. Unfortunately
it can also consume large ammounts of memory since each web server
maintains its own cache in its own memory space.
By using IPC::ShareLite and Storable (IPC::ShareLite and Storable),
HTML::Template was able to maintain a single shared cache of templates.
The downside was that HTML::Template's cache routines became
complicated by a lot of IPC code. My solution is to break out the IPC
cache mechanisms into their own module, IPC::SharedCache. Hopefully
over time it can become general enough to be usable by more than just
HTML::Template.
USAGE
This module allows you to store data in shared memory and have it load
automatically when needed. You can also define a test to screen cached
data for vailidty - if the test fails the data will be reloaded. This
is useful for defining a max-age for cached data or keeping cached data
in sync with other resources. In the web server example above the
validation test would look to see wether the file had changed on disk.
To initialize this module you provide two callback subroutines. The
first is the "load_callback". This gets called when a user of the
cache requests an item from that is not yet present or is stale. It
must return a reference to the data structure that will be stored in
the cache. The second is the "validate_callback". This gets called on
every cache access - its job is to check the cached object for
freshness (and/or some other validity, of course). It must return true
or false. When it returns true, the cached object is valid and is
retained in the cache. When it returns false, the object is re-loaded
using the "load_callback" and the result is stored in the cache.
To use the module you just request entries for the objects you need.
If the object is present in the cache and the "validate_callback"
returns true, then you get the object from the cache. If not, the
object is loaded into the cache with the "load_callback" and returned
to you.
The cache can be used to store any perl data structures that can be
serialized by the Storable module. See Storable for details.
EXAMPLE
In this example a shared cache of files is maintained. The
"load_callback" reads the file from disk into the cache and the
"validate_callback" checks its modification time using stat(). Note
that the "load_callback" stores information into the cached object that
"validate_callback" uses to check the freshness of the cache.
# the "load_callback", loads the file from disk, storing its stat()
# information along with the file into the cache. The key in this
# case is the filename to load.
sub load_file {
my $key = shift;
open(FILE, $key) or die "Unable to open file named $key : $!");
# note the modification time of this file - the 9th element of a
# stat() is the modification time of the file.
my $mtime = (stat($key))[9];
# read the file into the variable $contents in 1k chunks
my ($buffer, $contents);
while(read(FILE, $buffer, 1024)) { $contents .= $buffer }
close(FILE);
# prepare the record to store in the cache
my %record = ( mtime => $mtime, contents => $contents );
# this record goes into the cache associated with $key, which is
# the filename. Notice that we're returning a reference to the
# data structure here.
return \%record;
}
# the "validate" callback, checks the mtime of the file on disk and
# compares it to the cache value. The $record is a reference to the
# cached values array returned from load_file above.
sub validate_file {
my ($key, $record) = @_;
# get the modification time out of the record
my $stored_mtime = $record->{mtime};
# get the current modification time from the filesystem - the 9th
# element of a stat() is the modification time of the file.
my $current_mtime = (stat($key))[9];
# compare and return the appropriate result.
if ($stored_mtime == $current_mtime) {
# the cached object is valid, return true
return 1;
} else {
# the cached object is stale, return false - load_callback will
# be called to load it afresh from disk.
return 0;
}
}
# now we can construct the IPC::SharedCache object, using as a root
# key 'SAMS'.
tie %cache 'IPC::SharedCache' ipc_key => 'SAMS',
load_callback => \&load_file,
validate_callback => \&validate_file;
# fetch an object from the cache - if it's already in the cache and
# validate_file() returns 1, then we'll get the cached file. If it's
# not in the cache, or validate_file returns 0, then load_file is
# called to load the file into the cache.
$config_file = $cache{'/some/path/to/some.config'};
DETAILS
The module implements a full tied hash interface, meaning that you can
use exists(), delete(), keys() and each(). However, in normal usage
all you'll need to do is to fetch values from the cache and possible
delete keys. Just in case you were wondering, exists() doesn't trigger
a cache load - it returns 1 if the given key is already in the cache
and 0 if it isn't. Similarily, keys() and each() operate on key/value
pairs already loaded into the cache.
The most important thing to realize is that there is no need to
explicitely store into the cache since the load_callback is called
automatically when it is necessary to load new data. If you find
yourself using more than just ""$data = $cache{'key'};"" you need to
make sure you really know what you're doing!
OPTIONS
There are a number parameters to tie that can be used to control the
behavior of IPC::SharedCache. Some of them are required, and some art
optional. Here's a preview:
tie %cache, 'IPC::SharedCache',
# required parameters
ipc_key => 'MYKI',
load_callback => \&load,
validate_callback => \&validate,
# optional parameters
ipc_mode => 0666,
ipc_segment_size => 1_000_000,
max_size => 50_000_000,
debug => 1;
ipc_key (required)
This is the unique identifier for the particular cache. It can be
specified as either a four-character string or an integer value. Any
script that wishes to access the cache must use the same ipc_key value.
You can use the ftok() function from IPC::SysV to generate this value,
see IPC::SysV for details. Using an ipc_key value that's already in
use by a non-IPC::SharedCache application will cause an error. Many
systems provide a utility called 'ipcs' to examine shared memory; you
can use it to check for existing shared memory usage before choosing
your ipc_key.
load_callback and validate_callback (required)
These parameters both specify callbacks for IPC::SharedCache to use
when the cache gets a request for a key. When you access the cache
("$data = $cache{$key}"), the cache first looks to see if it already
has an object for the given key. If it doesn't, it calls the
load_callback and returns the result which is also stored in the cache.
Alternately, if it does have the object in the cache it calls the
validate_callback to check if the object is still good. If the
validate_callback returns true then object is good and is returned. If
the validate_callback returns false then the object is discarded and
the load_callback is called.
The load_callback recieves a single parameter - the requested key. It
must return a reference to the data object be stored in the cache.
Returning something that is not a reference results in an error.
The validate_callback recieves two parameters - the key and the
reference to the stored object. It must return true or false.
There are two ways to specify the callbacks. The first is simply to
specify a subroutine reference. This can be an anonymous subroutine or
a named one. Example:
tie %cache, 'IPC::SharedCache',
ipc_key => 'TEST',
load_callback => sub { ... },
validate_callback => \&validate;
The second method allows parameters to be passed to the subroutine when
it is called. This is done by specifying a reference to an array of
values, the first being the subroutine reference and the rest are
parameters for the subroutine. The extra parameters are passed in
before the IPC::SharedCache provided parameters. Example:
tie %cache, 'IPC::SharedCache',
ipc_key => 'TEST',
load_callback => [\&load, $arg1, $arg2, $arg3]
validate_callback => [\&validate, $self];
ipc_mode (optional)
This option specifies the access mode of the IPC cache. It defaults to
0666. See IPC::ShareLite for more information on IPC access modes.
The default should be fine for most applications.
ipc_segment_size (optional)
This option allows you to specify the "chunk size" of the IPC shared
memory segments. The default is 65,536, which is 64K. This is a good
default and is very portable. If you know that your system supports
larger IPC segment sizes and you know that your cache will be storing
large data items you might get better performance by increasing this
value.
This value places no limit on the size of an object stored in the cache
- IPC::SharedCache automatically spreads large objects across multiple
IPC segments.
WARNING: setting this value too low (below 1024 in my experience) can
cause errors.
max_size (optional)
By setting this parameter you are setting a logical maximum to the
ammount of data stored in the cache. When an item is stored in the
cache and this limit is exceded the oldest item (or items, as
necessary) in the cache will be deleted to make room. This value is
specified in bytes. It defaults to 0, which specifies no limit on the
size of the cache.
Turning this feature on costs a fair ammount of performance - how much
depends largely on home much data is being stored into the cache versus
the size of max_cache. In the worst case (where the max_size is set
much too low) this option can cause severe "thrashing" and negate the
benefit of maintaining a cache entirely.
NOTE: The size of the cache may in fact exceed this value - the book-
keeping data stored in the root segment is not counted towards the
total. Also, extra padding imposed by the ipc_segment_size is not
counted. This may change in the future if I learn that it would be
appropriate to count this padding as used memory. It is not clear to
me that all IPC implementations will really waste this memory.
debug (optional)
Set this option to 1 to see a whole bunch of text on STDERR about what
IPC::SharedCache is doing.
UTILITIES
Two static functions are included in this package that are meant to be
used from the command-line.
walk
Walk prints out a detailed listing of the contents of a shared cache at
a given ipc_key. It provides information the current keys stored and a
dump of the objects stored in each key. Be warned, this can be quite a
lot of data! Also, you'll need the Data::Dumper module installed to
use 'walk'. You can get it on CPAN.
You can call walk like:
perl -MIPC::SharedCache -e 'IPC::SharedCache::walk AKEY'"
Example:
$ perl -MIPC::SharedCache -e 'IPC::SharedCache::walk MYKI'"
*===================*
IPC::SharedCache Root
*===================*
IPC_KEY: MYKI
ELEMENTS: 3
TOTAL SIZE: 99 bytes
KEYS: a, b, c
*=======*
Data List
*=======*
KEY: a
$CONTENTS = [
950760892,
950760892,
950760892
];
KEY: b
$CONTENTS = [
950760892,
950760892,
950760892
];
KEY: c
$CONTENTS = [
950760892,
950760892,
950760892
];
remove
This function totally removes an entire cache given an ipc_key value.
This should not be done to a running system! Still, it's an invaluable
tool during development when flawed data may become 'stuck' in the
cache.
$ perl -MIPC::SharedCache -e 'IPC::SharedCache::remove MYKI'
This function is silent and thus may be usefully called from within a
script if desired.
BUGS
I am aware of no bugs - if you find one please email me at
sam@tregar.com. When submitting bug reports, be sure to include full
details, including the VERSION of the module and a test script
demonstrating the problem.
CREDITS
I would like to thank Maurice Aubrey for making this module possible by
producing the excelent IPC::ShareLite.
The following people have contributed patches, ideas or new features:
Tim Bunce
Roland Mas
Drew Taylor
Ed Loehr
Maverick
Thanks everyone!
AUTHOR
Sam Tregar, sam@tregar.com (you can also find me on the mailing list
for HTML::Template at htmltmpl@lists.vm.com - join it by sending a
blank message to htmltmpl-subscribe@lists.vm.com).
LICENSEIPC::SharedCache - a Perl module to manage a SysV IPC shared cache.
Copyright (C) 2000 Sam Tregar (sam@tregar.com)
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
AUTHOR
Sam Tregar, sam@tregar.com
SEE ALSOperl(1).
perl v5.14.1 2011-06-20 SharedCache(3)