Caching Pages One Way or Another

The Different ways I Can Cache You Page

Overview

Page caching can happen at application level via a cache plugin such as Catalyst::Plugin::Cache with a backend such as Cache::FastMmap. It can also be done on the web server or via a proxy. This paper will examine PageCache Catalyst plugin, NGINX web server and Varnish Proxy cache control techniques.

It’s important to highlight the facets of caching. On one hand, deciding what to put in the cache (and where to store it). While another other hand concerns with when to pull (read) from the cache (and where to get it from.

• set cache
• get cache

Set Cache

We need to set the cache before we can get anything beside nothing. The cache can actually be set at different places in the application.

• PageCache Catalyst plugin
• NGINX expires directive in configuration
• Varnish VCL

Further yet, setting a cache may be putting an item into a cache or setting the header of the item perform Cache-Control.

PageCache

Backend Cache

One way to cache MojoMojo is to do so at the application level. One specifies which pages to cache and for how long they should be in the cache.

Headers for Caching Middleware

One can also set some headers for Cache-Control such as max-age or no-cache

Application Cache Configuration

use Catalyst qw/
    # Other Modules Before
    PageCache
    /;

MojoMojo->config(
        'Plugin::PageCache' => {
            expires          => 300,
            set_http_headers => 0,
            auto_check_user  => 1,
            auto_cache       => [
                               '/.*',
            ],
            key_maker => sub {
                my $c = shift;
                return $c->stash->{path} . '.' . $c->req->path;
            },
            debug => 0 ,
            cache_hook => 'cache_hook'
        }
);

Notes about cache configuration keys:

• expires - how long a page is cached before being rebuilt
• set_http_headers - can set Cache-Control header such as ‘max-age’
• auto_check_user - don’t cache pages when a user is logged in¹
• auto_cache - specifies uri’s to automatically cache (takes regex)
• key_maker - over method to create unique key for cache items
• debug - show if a page is in the cache (needs Catalyst Debug on, -d)
• cache_hook - a callback to decide whether a page should be cached or not

where the cache_hook is a method name that determines (boolean) whether a page is to be cached or not. Thus this is a application wide approach.

cache_hook

The guidelines for the cache_hook are to cache unless

• $c->user_exists (auto_check_user)
• CATALYST_NOCACHE is set
• path is excluded from cache_ie_list

indirect method

One cache_hook method for MojoMojo currently looks like:

sub cache_hook {
  my ( $c ) = @_;

  if ( $c->user_exists        
    || $ENV{CATALYST_NOCACHE} 
    || ! $c->cache_ie_list->evaluate($c->req->path)
     ) {
    return 0; # Don't cache
  }
  return 1;   # Cache
}

Notice we don’t cache when user exists (similar if not the same as auto_check_user) nor when and environment variable has been set which can turn off PageCache all-togther.

direct method

sub cache_hook {
  my ( $c ) = @_;

  return !$c->user_exists && !$ENV{CATALYST_NOCACHE} && $c->cache_ie_list->evaluate($c->req->path);

}

cache_ie_list

URI’s are either included or excluded from the cache.

my $ie;
$ie = Algorithm::IncludeExclude->new;

# Start by including all page for caching
$ie->include(); 
# static files will be handled via web server or proxy cache control.
$ie->exclude(qr/static/);
$ie->exclude('login');
$ie->exclude('logout');
$ie->exclude('edit');
#$ie->exclude('list');
#$ie->exclude('recent');

sub cache_ie_list {
    return $ie;
}

Cache in Controller

Besides application wide controller, one can cache at the level of a controller action ala:

sub view : Global {
    my ( $self, $c, $path ) = @_;
    
    $c->cache_page(
        last_modified   => $last_modified,
        cache_seconds   => 24 * 60 * 60,
        expires         => 30,
    ); 

which separates control of page cache storage time and the header expiration date (which can be used by proxies).

Firebug Info

Let’s see what information (headers in particular) Firebug in FireFox will show us.

Catalyst Built-In Server - set_http_headers => 0 (OFF)

Capçaleres de la resposta

X-PageCache	Catalyst

Capçaleres de la sol·licitud

Cookie          mojomojo_session=43b7b9dc41f21d7219051d2343d8cc66b6ca9a48; split_edit=1

Catalyst Built-In Server - set_http_headers => 1 (ON)

Capçaleres de la resposta

Cache-Control	max-age=251
X-PageCache	Catalyst

Capçaleres de la sol·licitud

Cookie          mojomojo_session=43b7b9dc41f21d7219051d2343d8cc66b6ca9a48; split_edit=1

NGINX Server - set_http_headers => 0 (OFF)

Capçaleres de la resposta

Server	        nginx/0.6.32
X-PageCache	Catalyst

Capçaleres de la sol·licitud

Cookie          mojomojo_session=43b7b9dc41f21d7219051d2343d8cc66b6ca9a48; split_edit=1

NGINX Server - set_http_headers => 1 (ON)

Capçaleres de la resposta

Server	        nginx/0.6.32
Cache-Control	max-age=251
X-PageCache	Catalyst

Capçaleres de la sol·licitud

Cookie          mojomojo_session=43b7b9dc41f21d7219051d2343d8cc66b6ca9a48; split_edit=1

¹ This doesn’t prevent a logged in user from receiving a cached page. It just that they won’t cause any pages to be cached.

My tags:

 

Popular tags: