1 <?xml version="1.0" encoding="UTF-8" ?>
2 <!DOCTYPE manualpage SYSTEM "style/manualpage.dtd">
3 <?xml-stylesheet type="text/xsl" href="style/manual.en.xsl"?>
4 <!-- $LastChangedRevision$ -->
7 Licensed to the Apache Software Foundation (ASF) under one or more
8 contributor license agreements. See the NOTICE file distributed with
9 this work for additional information regarding copyright ownership.
10 The ASF licenses this file to You under the Apache License, Version 2.0
11 (the "License"); you may not use this file except in compliance with
12 the License. You may obtain a copy of the License at
14 http://www.apache.org/licenses/LICENSE-2.0
16 Unless required by applicable law or agreed to in writing, software
17 distributed under the License is distributed on an "AS IS" BASIS,
18 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
19 See the License for the specific language governing permissions and
20 limitations under the License.
23 <manualpage metafile="caching.xml.meta">
25 <title>Caching Guide</title>
28 <p>This document supplements the <module>mod_cache</module>,
29 <module>mod_cache_disk</module>, <module>mod_file_cache</module> and <a
30 href="programs/htcacheclean.html">htcacheclean</a> reference documentation.
31 It describes how to use the Apache HTTP Server's caching features to accelerate web and
32 proxy serving, while avoiding common problems and misconfigurations.</p>
35 <section id="introduction">
36 <title>Introduction</title>
38 <p>The Apache HTTP server offers a range of caching features that
39 are designed to improve the performance of the server in various
43 <dt>Three-state RFC2616 HTTP caching</dt>
45 <module>mod_cache</module>
46 and its provider modules
47 <module>mod_cache_disk</module>
48 provide intelligent, HTTP-aware caching. The content itself is stored
49 in the cache, and mod_cache aims to honor all of the various HTTP
50 headers and options that control the cacheability of content
52 <a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html">Section
54 <module>mod_cache</module>
55 is aimed at both simple and complex caching configurations, where
56 you are dealing with proxied content, dynamic local content or
57 have a need to speed up access to local files on a potentially
61 <dt>Two-state key/value shared object caching</dt>
63 The <a href="socache.html">shared object cache API</a> (socache)
64 and its provider modules provide a
65 server wide key/value based shared object cache. These modules
66 are designed to cache low level data such as SSL sessions and
67 authentication credentials. Backends allow the data to be stored
68 server wide in shared memory, or datacenter wide in a cache such
69 as memcache or distcache.
72 <dt>Specialized file caching</dt>
74 <module>mod_file_cache</module>
75 offers the ability to pre-load
76 files into memory on server startup, and can improve access
77 times and save file handles on files that are accessed often,
78 as there is no need to go to disk on each request.
82 <p>To get the most from this document, you should be familiar with
83 the basics of HTTP, and have read the Users' Guides to
84 <a href="urlmapping.html">Mapping URLs to the Filesystem</a> and
85 <a href="content-negotiation.html">Content negotiation</a>.</p>
89 <section id="http-caching">
91 <title>Three-state RFC2616 HTTP caching</title>
95 <module>mod_cache</module>
96 <module>mod_cache_disk</module>
99 <directive module="mod_cache">CacheEnable</directive>
100 <directive module="mod_cache">CacheDisable</directive>
101 <directive module="core">UseCanonicalName</directive>
102 <directive module="mod_negotiation">CacheNegotiatedDocs</directive>
106 <p>The HTTP protocol contains built in support for an in-line caching
108 <a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html">
109 described by section 13 of RFC2616</a>, and the
110 <module>mod_cache</module> module can be used to take advantage of
113 <p>Unlike a simple two state key/value cache where the content
114 disappears completely when no longer fresh, an HTTP cache includes
115 a mechanism to retain stale content, and to ask the origin server
116 whether this stale content has changed and if not, make it fresh
119 <p>An entry in an HTTP cache exists in one of three states:</p>
124 If the content is new enough (younger than its <strong>freshness
125 lifetime</strong>), it is considered <strong>fresh</strong>. An
126 HTTP cache is free to serve fresh content without making any
127 calls to the origin server at all.
131 <p>If the content is too old (older than its <strong>freshness
132 lifetime</strong>), it is considered <strong>stale</strong>. An
133 HTTP cache should contact the origin server and check whether
134 the content is still fresh before serving stale content to a
135 client. The origin server will either respond with replacement
136 content if not still valid, or ideally, the origin server will
137 respond with a code to tell the cache the content is still
138 fresh, without the need to generate or send the content again.
139 The content becomes fresh again and the cycle continues.</p>
141 <p>The HTTP protocol does allow the cache to serve stale data
142 under certain circumstances, such as when an attempt to freshen
143 the data with an origin server has failed with a 5xx error, or
144 when another request is already in the process of freshening
145 the given entry. In these cases a <code>Warning</code> header
146 is added to the response.</p>
148 <dt>Non Existent</dt>
150 If the cache gets full, it reserves the option to delete content
151 from the cache to make space. Content can be deleted at any time,
152 and can be stale or fresh. The <a
153 href="programs/htcacheclean.html">htcacheclean</a> tool can be
154 run on a once off basis, or deployed as a daemon to keep the size
155 of the cache within the given size, or the given number of inodes.
156 The tool attempts to delete stale content before attempting to
157 delete fresh content.
161 <p>Full details of how HTTP caching works can be found in
162 <a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html">
163 Section 13 of RFC2616</a>.</p>
166 <title>Interaction with the Server</title>
168 <p>The <module>mod_cache</module> module hooks into the server in two
169 possible places depending on the value of the
170 <directive module="mod_cache">CacheQuickHandler</directive> directive:
174 <dt>Quick handler phase</dt>
176 <p>This phase happens very early on during the request processing,
177 just after the request has been parsed. If the content is
178 found within the cache, it is served immediately and almost
179 all request processing is bypassed.</p>
181 <p>In this scenario, the cache behaves as if it has been "bolted
182 on" to the front of the server.</p>
184 <p>This mode offers the best performance, as the majority of
185 server processing is bypassed. This mode however also bypasses the
186 authentication and authorization phases of server processing, so
187 this mode should be chosen with care when this is important.</p>
189 <p> Requests with an "Authorization" header (for example, HTTP Basic
190 Authentication) are neither cacheable nor served from the cache
191 when <module>mod_cache</module> is running in this phase.</p>
193 <dt>Normal handler phase</dt>
195 <p>This phase happens late in the request processing, after all
196 the request phases have completed.</p>
198 <p>In this scenario, the cache behaves as if it has been "bolted
199 on" to the back of the server.</p>
201 <p>This mode offers the most flexibility, as the potential exists
202 for caching to occur at a precisely controlled point in the filter
203 chain, and cached content can be filtered or personalized before
204 being sent to the client.</p>
208 <p>If the URL is not found within the cache, <module>mod_cache</module>
209 will add a <a href="filter.html">filter</a> to the filter stack in order
210 to record the response to the cache, and then stand down, allowing normal
211 request processing to continue. If the content is determined to be
212 cacheable, the content will be saved to the cache for future serving,
213 otherwise the content will be ignored.</p>
215 <p>If the content found within the cache is stale, the
216 <module>mod_cache</module> module converts the request into a
217 <strong>conditional request</strong>. If the origin server responds with
218 a normal response, the normal response is cached, replacing the content
219 already cached. If the origin server responds with a 304 Not Modified
220 response, the content is marked as fresh again, and the cached content
221 is served by the filter instead of saving it.</p>
225 <title>Improving Cache Hits</title>
227 <p>When a virtual host is known by one of many different server aliases,
228 ensuring that <directive module="core">UseCanonicalName</directive> is
229 set to <code>On</code> can dramatically improve the ratio of cache hits.
230 This is because the hostname of the virtual-host serving the content is
231 used within the cache key. With the setting set to <code>On</code>
232 virtual-hosts with multiple server names or aliases will not produce
233 differently cached entities, and instead content will be cached as
234 per the canonical hostname.</p>
239 <title>Freshness Lifetime</title>
241 <p>Well formed content that is intended to be cached should declare an
242 explicit freshness lifetime with the <code>Cache-Control</code>
243 header's <code>max-age</code> or <code>s-maxage</code> fields, or
244 by including an <code>Expires</code> header.</p>
246 <p>At the same time, the origin server defined freshness lifetime can
247 be overridden by a client when the client presents their own
248 <code>Cache-Control</code> header within the request. In this case,
249 the lowest freshness lifetime between request and response wins.</p>
251 <p>When this freshness lifetime is missing from the request or the
252 response, a default freshness lifetime is applied. The default
253 freshness lifetime for cached entities is one hour, however
254 this can be easily over-ridden by using the <directive
255 module="mod_cache">CacheDefaultExpire</directive> directive.</p>
257 <p>If a response does not include an <code>Expires</code> header but does
258 include a <code>Last-Modified</code> header, <module>mod_cache</module>
259 can infer a freshness lifetime based on a heuristic, which can be
260 controlled through the use of the <directive
261 module="mod_cache">CacheLastModifiedFactor</directive> directive.</p>
263 <p>For local content, or for remote content that does not define its own
264 <code>Expires</code> header, <module>mod_expires</module> may be used to
265 fine-tune the freshness lifetime by adding <code>max-age</code> and
266 <code>Expires</code>.</p>
268 <p>The maximum freshness lifetime may also be controlled by using the
269 <directive module="mod_cache">CacheMaxExpire</directive>.</p>
274 <title>A Brief Guide to Conditional Requests</title>
276 <p>When content expires from the cache and becomes stale, rather than
277 pass on the original request, httpd will modify the request to make
278 it conditional instead.</p>
280 <p>When an <code>ETag</code> header exists in the original cached
281 response, <module>mod_cache</module> will add an
282 <code>If-None-Match</code> header to the request to the origin server.
283 When a <code>Last-Modified</code> header exists in the original
284 cached response, <module>mod_cache</module> will add an
285 <code>If-Modified-Since</code> header to the request to the origin
286 server. Performing either of these actions makes the request
287 <strong>conditional</strong>.</p>
289 <p>When a conditional request is received by an origin server, the
290 origin server should check whether the ETag or the Last-Modified
291 parameter has changed, as appropriate for the request. If not, the
292 origin should respond with a terse "304 Not Modified" response. This
293 signals to the cache that the stale content is still fresh should be
294 used for subsequent requests until the content's new freshness lifetime
295 is reached again.</p>
297 <p>If the content has changed, then the content is served as if the
298 request were not conditional to begin with.</p>
300 <p>Conditional requests offer two benefits. Firstly, when making such
301 a request to the origin server, if the content from the origin
302 matches the content in the cache, this can be determined easily and
303 without the overhead of transferring the entire resource.</p>
305 <p>Secondly, a well designed origin server will be designed in such
306 a way that conditional requests will be significantly cheaper to
307 produce than a full response. For static files, typically all that is
308 involved is a call to <code>stat()</code> or similar system call, to
309 see if the file has changed in size or modification time. As such, even
310 local content may still be served faster from the cache if it has not
313 <p>Origin servers should make every effort to support conditional
314 requests as is practical, however if conditional requests are not
315 supported, the origin will respond as if the request was not
316 conditional, and the cache will respond as if the content had changed
317 and save the new content to the cache. In this case, the cache will
318 behave like a simple two state cache, where content is effectively
319 either fresh or deleted.</p>
323 <title>What Can be Cached?</title>
325 <p>The full definition of which responses can be cached by an HTTP
327 <a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html#sec13.4">
328 RFC2616 Section 13.4 Response Cacheability</a>, and can be summed up as
332 <li>Caching must be enabled for this URL. See the <directive
333 module="mod_cache">CacheEnable</directive> and <directive
334 module="mod_cache">CacheDisable</directive> directives.</li>
336 <li>The response must have a HTTP status code of 200, 203, 300, 301 or
339 <li>The request must be a HTTP GET request.</li>
341 <li>If the response contains an "Authorization:" header, it must
342 also contain an "s-maxage", "must-revalidate" or "public" option
343 in the "Cache-Control:" header, or it won't be cached.</li>
345 <li>If the URL included a query string (e.g. from a HTML form GET
346 method) it will not be cached unless the response specifies an
347 explicit expiration by including an "Expires:" header or the max-age
348 or s-maxage directive of the "Cache-Control:" header, as per RFC2616
349 sections 13.9 and 13.2.1.</li>
351 <li>If the response has a status of 200 (OK), the response must
352 also include at least one of the "Etag", "Last-Modified" or
353 the "Expires" headers, or the max-age or s-maxage directive of
354 the "Cache-Control:" header, unless the
355 <directive module="mod_cache">CacheIgnoreNoLastMod</directive>
356 directive has been used to require otherwise.</li>
358 <li>If the response includes the "private" option in a "Cache-Control:"
359 header, it will not be stored unless the
360 <directive module="mod_cache">CacheStorePrivate</directive> has been
361 used to require otherwise.</li>
363 <li>Likewise, if the response includes the "no-store" option in a
364 "Cache-Control:" header, it will not be stored unless the
365 <directive module="mod_cache">CacheStoreNoStore</directive> has been
368 <li>A response will not be stored if it includes a "Vary:" header
369 containing the match-all "*".</li>
374 <title>What Should Not be Cached?</title>
376 <p>It should be up to the client creating the request, or the origin
377 server constructing the response to decide whether or not the content
378 should be cacheable or not by correctly setting the
379 <code>Cache-Control</code> header, and <module>mod_cache</module> should
380 be left alone to honor the wishes of the client or server as appropriate.
383 <p>Content that is time sensitive, or which varies depending on the
384 particulars of the request that are not covered by HTTP negotiation,
385 should not be cached. This content should declare itself uncacheable
386 using the <code>Cache-Control</code> header.</p>
388 <p>If content changes often, expressed by a freshness lifetime of minutes
389 or seconds, the content can still be cached, however it is highly
390 desirable that the origin server supports
391 <strong>conditional requests</strong> correctly to ensure that
392 full responses do not have to be generated on a regular basis.</p>
394 <p>Content that varies based on client provided request headers can be
395 cached through intelligent use of the <code>Vary</code> response
401 <title>Variable/Negotiated Content</title>
403 <p>When the origin server is designed to respond with different content
404 based on the value of headers in the request, for example to serve
405 multiple languages at the same URL, HTTP's caching mechanism makes it
406 possible to cache multiple variants of the same page at the same URL.</p>
408 <p>This is done by the origin server adding a <code>Vary</code> header
409 to indicate which headers must be taken into account by a cache when
410 determining whether two variants are different from one another.</p>
412 <p>If for example, a response is received with a vary header such as;</p>
415 Vary: negotiate,accept-language,accept-charset
418 <p><module>mod_cache</module> will only serve the cached content to
419 requesters with accept-language and accept-charset headers
420 matching those of the original request.</p>
422 <p>Multiple variants of the content can be cached side by side,
423 <module>mod_cache</module> uses the <code>Vary</code> header and the
424 corresponding values of the request headers listed by <code>Vary</code>
425 to decide on which of many variants to return to the client.</p>
429 <title>Caching to Disk</title>
431 <p>The <module>mod_cache</module> module relies on specific backend store
432 implementations in order to manage the cache, and for caching to disk
433 <module>mod_cache_disk</module> is provided to support this.</p>
435 <p>Typically the module will be configured as so;</p>
437 <highlight language="config">
438 CacheRoot "/var/cache/apache/"
444 <p>Importantly, as the cached files are locally stored, operating system
445 in-memory caching will typically be applied to their access also. So
446 although the files are stored on disk, if they are frequently accessed
447 it is likely the operating system will ensure that they are actually
448 served from memory.</p>
453 <title>Understanding the Cache-Store</title>
455 <p>To store items in the cache, <module>mod_cache_disk</module> creates
456 a 22 character hash of the URL being requested. This hash incorporates
457 the hostname, protocol, port, path and any CGI arguments to the URL,
458 as well as elements defined by the Vary header to ensure that multiple
459 URLs do not collide with one another.</p>
461 <p>Each character may be any one of 64-different characters, which mean
462 that overall there are 64^22 possible hashes. For example, a URL might
463 be hashed to <code>xyTGxSMO2b68mBCykqkp1w</code>. This hash is used
464 as a prefix for the naming of the files specific to that URL within
465 the cache, however first it is split up into directories as per
466 the <directive module="mod_cache_disk">CacheDirLevels</directive> and
467 <directive module="mod_cache_disk">CacheDirLength</directive>
470 <p><directive module="mod_cache_disk">CacheDirLevels</directive>
471 specifies how many levels of subdirectory there should be, and
472 <directive module="mod_cache_disk">CacheDirLength</directive>
473 specifies how many characters should be in each directory. With
474 the example settings given above, the hash would be turned into
476 <code>/var/cache/apache/x/y/TGxSMO2b68mBCykqkp1w</code>.</p>
478 <p>The overall aim of this technique is to reduce the number of
479 subdirectories or files that may be in a particular directory,
480 as most file-systems slow down as this number increases. With
482 <directive module="mod_cache_disk">CacheDirLength</directive>
483 there can at most be 64 subdirectories at any particular level.
484 With a setting of 2 there can be 64 * 64 subdirectories, and so on.
485 Unless you have a good reason not to, using a setting of "1"
486 for <directive module="mod_cache_disk">CacheDirLength</directive>
490 <directive module="mod_cache_disk">CacheDirLevels</directive>
491 depends on how many files you anticipate to store in the cache.
492 With the setting of "2" used in the above example, a grand
493 total of 4096 subdirectories can ultimately be created. With
494 1 million files cached, this works out at roughly 245 cached
495 URLs per directory.</p>
497 <p>Each URL uses at least two files in the cache-store. Typically
498 there is a ".header" file, which includes meta-information about
499 the URL, such as when it is due to expire and a ".data" file
500 which is a verbatim copy of the content to be served.</p>
502 <p>In the case of a content negotiated via the "Vary" header, a
503 ".vary" directory will be created for the URL in question. This
504 directory will have multiple ".data" files corresponding to the
505 differently negotiated content.</p>
509 <title>Maintaining the Disk Cache</title>
511 <p>The <module>mod_cache_disk</module> module makes no attempt to
512 regulate the amount of disk space used by the cache, although it
513 will gracefully stand down on any disk error and behave as if the
514 cache was never present.</p>
516 <p>Instead, provided with httpd is the <a
517 href="programs/htcacheclean.html">htcacheclean</a> tool which allows you
518 to clean the cache periodically. Determining how frequently to run <a
519 href="programs/htcacheclean.html">htcacheclean</a> and what target size to
520 use for the cache is somewhat complex and trial and error may be needed to
521 select optimal values.</p>
523 <p><a href="programs/htcacheclean.html">htcacheclean</a> has two modes of
524 operation. It can be run as persistent daemon, or periodically from
526 href="programs/htcacheclean.html">htcacheclean</a> can take up to an hour
527 or more to process very large (tens of gigabytes) caches and if you are
528 running it from cron it is recommended that you determine how long a typical
529 run takes, to avoid running more than one instance at a time.</p>
531 <p>It is also recommended that an appropriate "nice" level is chosen for
532 htcacheclean so that the tool does not cause excessive disk io while the
533 server is running.</p>
536 <img src="images/caching_fig1.gif" alt="" width="600"
537 height="406" /><br />
538 <a id="figure1" name="figure1"><dfn>Figure 1</dfn></a>: Typical
539 cache growth / clean sequence.</p>
541 <p>Because <module>mod_cache_disk</module> does not itself pay attention
542 to how much space is used you should ensure that
543 <a href="programs/htcacheclean.html">htcacheclean</a> is configured to
544 leave enough "grow room" following a clean.</p>
549 <section id="socache-caching">
551 <title>Two-state Key/Value Shared Object Caching</title>
555 <module>mod_authn_socache</module>
556 <module>mod_socache_dbm</module>
557 <module>mod_socache_dc</module>
558 <module>mod_socache_memcache</module>
559 <module>mod_socache_shmcb</module>
560 <module>mod_ssl</module>
563 <directive module="mod_authn_socache">AuthnCacheSOCache</directive>
564 <directive module="mod_ssl">SSLSessionCache</directive>
565 <directive module="mod_ssl">SSLStaplingCache</directive>
569 <p>The Apache HTTP server offers a low level shared object cache for
570 caching information such as SSL sessions, or authentication credentials,
571 within the <a href="socache.html">socache</a> interface.</p>
573 <p>Additional modules are provided for each implementation, offering the
574 following backends:</p>
577 <dt><module>mod_socache_dbm</module></dt>
578 <dd>DBM based shared object cache.</dd>
579 <dt><module>mod_socache_dc</module></dt>
580 <dd>Distcache based shared object cache.</dd>
581 <dt><module>mod_socache_memcache</module></dt>
582 <dd>Memcache based shared object cache.</dd>
583 <dt><module>mod_socache_shmcb</module></dt>
584 <dd>Shared memory based shared object cache.</dd>
587 <section id="mod_authn_socache-caching">
588 <title>Caching Authentication Credentials</title>
592 <module>mod_authn_socache</module>
595 <directive module="mod_authn_socache">AuthnCacheSOCache</directive>
599 <p>The <module>mod_authn_socache</module> module allows the result of
600 authentication to be cached, relieving load on authentication backends.</p>
604 <section id="mod_ssl-caching">
605 <title>Caching SSL Sessions</title>
609 <module>mod_ssl</module>
612 <directive module="mod_ssl">SSLSessionCache</directive>
613 <directive module="mod_ssl">SSLStaplingCache</directive>
617 <p>The <module>mod_ssl</module> module uses the <code>socache</code> interface
618 to provide a session cache and a stapling cache.</p>
624 <section id="file-caching">
626 <title>Specialized File Caching</title>
630 <module>mod_file_cache</module>
633 <directive module="mod_file_cache">CacheFile</directive>
634 <directive module="mod_file_cache">MMapFile</directive>
638 <p>On platforms where a filesystem might be slow, or where file
639 handles are expensive, the option exists to pre-load files into
640 memory on startup.</p>
642 <p>On systems where opening files is slow, the option exists to
643 open the file on startup and cache the file handle. These
644 options can help on systems where access to static files is
647 <section id="filehandle">
648 <title>File-Handle Caching</title>
650 <p>The act of opening a file can itself be a source of delay, particularly
651 on network filesystems. By maintaining a cache of open file descriptors
652 for commonly served files, httpd can avoid this delay. Currently httpd
653 provides one implementation of File-Handle Caching.</p>
656 <title>CacheFile</title>
658 <p>The most basic form of caching present in httpd is the file-handle
659 caching provided by <module>mod_file_cache</module>. Rather than caching
660 file-contents, this cache maintains a table of open file descriptors. Files
661 to be cached in this manner are specified in the configuration file using
662 the <directive module="mod_file_cache">CacheFile</directive>
666 <directive module="mod_file_cache">CacheFile</directive> directive
667 instructs httpd to open the file when it is started and to re-use
668 this file-handle for all subsequent access to this file.</p>
670 <highlight language="config">
671 CacheFile /usr/local/apache2/htdocs/index.html
674 <p>If you intend to cache a large number of files in this manner, you
675 must ensure that your operating system's limit for the number of open
676 files is set appropriately.</p>
678 <p>Although using <directive module="mod_file_cache">CacheFile</directive>
679 does not cause the file-contents to be cached per-se, it does mean
680 that if the file changes while httpd is running these changes will
681 not be picked up. The file will be consistently served as it was
682 when httpd was started.</p>
684 <p>If the file is removed while httpd is running, it will continue
685 to maintain an open file descriptor and serve the file as it was when
686 httpd was started. This usually also means that although the file
687 will have been deleted, and not show up on the filesystem, extra free
688 space will not be recovered until httpd is stopped and the file
689 descriptor closed.</p>
694 <section id="inmemory">
695 <title>In-Memory Caching</title>
697 <p>Serving directly from system memory is universally the fastest method
698 of serving content. Reading files from a disk controller or, even worse,
699 from a remote network is orders of magnitude slower. Disk controllers
700 usually involve physical processes, and network access is limited by
701 your available bandwidth. Memory access on the other hand can take mere
704 <p>System memory isn't cheap though, byte for byte it's by far the most
705 expensive type of storage and it's important to ensure that it is used
706 efficiently. By caching files in memory you decrease the amount of
707 memory available on the system. As we'll see, in the case of operating
708 system caching, this is not so much of an issue, but when using
709 httpd's own in-memory caching it is important to make sure that you
710 do not allocate too much memory to a cache. Otherwise the system
711 will be forced to swap out memory, which will likely degrade
715 <title>Operating System Caching</title>
717 <p>Almost all modern operating systems cache file-data in memory managed
718 directly by the kernel. This is a powerful feature, and for the most
719 part operating systems get it right. For example, on Linux, let's look at
720 the difference in the time it takes to read a file for the first time
721 and the second time;</p>
724 colm@coroebus:~$ time cat testfile > /dev/null
728 colm@coroebus:~$ time cat testfile > /dev/null
734 <p>Even for this small file, there is a huge difference in the amount
735 of time it takes to read the file. This is because the kernel has cached
736 the file contents in memory.</p>
738 <p>By ensuring there is "spare" memory on your system, you can ensure
739 that more and more file-contents will be stored in this cache. This
740 can be a very efficient means of in-memory caching, and involves no
741 extra configuration of httpd at all.</p>
743 <p>Additionally, because the operating system knows when files are
744 deleted or modified, it can automatically remove file contents from the
745 cache when necessary. This is a big advantage over httpd's in-memory
746 caching which has no way of knowing when a file has changed.</p>
749 <p>Despite the performance and advantages of automatic operating system
750 caching there are some circumstances in which in-memory caching may be
751 better performed by httpd.</p>
754 <title>MMapFile Caching</title>
756 <p><module>mod_file_cache</module> provides the
757 <directive module="mod_file_cache">MMapFile</directive> directive, which
758 allows you to have httpd map a static file's contents into memory at
759 start time (using the mmap system call). httpd will use the in-memory
760 contents for all subsequent accesses to this file.</p>
762 <highlight language="config">
763 MMapFile /usr/local/apache2/htdocs/index.html
767 <directive module="mod_file_cache">CacheFile</directive> directive, any
768 changes in these files will not be picked up by httpd after it has
771 <p> The <directive module="mod_file_cache">MMapFile</directive>
772 directive does not keep track of how much memory it allocates, so
773 you must ensure not to over-use the directive. Each httpd child
774 process will replicate this memory, so it is critically important
775 to ensure that the files mapped are not so large as to cause the
776 system to swap memory.</p>
782 <section id="security">
783 <title>Security Considerations</title>
786 <title>Authorization and Access Control</title>
788 <p>Using <module>mod_cache</module> in its default state where
789 <directive module="mod_cache">CacheQuickHandler</directive> is set to
790 <code>On</code> is very much like having a caching reverse-proxy bolted
791 to the front of the server. Requests will be served by the caching module
792 unless it determines that the origin server should be queried just as an
793 external cache would, and this drastically changes the security model of
796 <p>As traversing a filesystem hierarchy to examine potential
797 <code>.htaccess</code> files would be a very expensive operation,
798 partially defeating the point of caching (to speed up requests),
799 <module>mod_cache</module> makes no decision about whether a cached
800 entity is authorised for serving. In other words; if
801 <module>mod_cache</module> has cached some content, it will be served
802 from the cache as long as that content has not expired.</p>
804 <p>If, for example, your configuration permits access to a resource by IP
805 address you should ensure that this content is not cached. You can do this
806 by using the <directive module="mod_cache">CacheDisable</directive>
807 directive, or <module>mod_expires</module>. Left unchecked,
808 <module>mod_cache</module> - very much like a reverse proxy - would cache
809 the content when served and then serve it to any client, on any IP
812 <p>When the <directive module="mod_cache">CacheQuickHandler</directive>
813 directive is set to <code>Off</code>, the full set of request processing
814 phases are executed and the security model remains unchanged.</p>
818 <title>Local exploits</title>
820 <p>As requests to end-users can be served from the cache, the cache
821 itself can become a target for those wishing to deface or interfere with
822 content. It is important to bear in mind that the cache must at all
823 times be writable by the user which httpd is running as. This is in
824 stark contrast to the usually recommended situation of maintaining
825 all content unwritable by the Apache user.</p>
827 <p>If the Apache user is compromised, for example through a flaw in
828 a CGI process, it is possible that the cache may be targeted. When
829 using <module>mod_cache_disk</module>, it is relatively easy to
830 insert or modify a cached entity.</p>
832 <p>This presents a somewhat elevated risk in comparison to the other
833 types of attack it is possible to make as the Apache user. If you are
834 using <module>mod_cache_disk</module> you should bear this in mind -
835 ensure you upgrade httpd when security upgrades are announced and
836 run CGI processes as a non-Apache user using <a
837 href="suexec.html">suEXEC</a> if possible.</p>
842 <title>Cache Poisoning</title>
844 <p>When running httpd as a caching proxy server, there is also the
845 potential for so-called cache poisoning. Cache Poisoning is a broad
846 term for attacks in which an attacker causes the proxy server to
847 retrieve incorrect (and usually undesirable) content from the origin
850 <p>For example if the DNS servers used by your system running httpd
851 are vulnerable to DNS cache poisoning, an attacker may be able to control
852 where httpd connects to when requesting content from the origin server.
853 Another example is so-called HTTP request-smuggling attacks.</p>
855 <p>This document is not the correct place for an in-depth discussion
856 of HTTP request smuggling (instead, try your favourite search engine)
857 however it is important to be aware that it is possible to make
858 a series of requests, and to exploit a vulnerability on an origin
859 webserver such that the attacker can entirely control the content
860 retrieved by the proxy.</p>
864 <title>Denial of Service / Cachebusting</title>
866 <p>The Vary mechanism allows multiple variants of the same URL to be
867 cached side by side. Depending on header values provided by the client,
868 the cache will select the correct variant to return to the client. This
869 mechanism can become a problem when an attempt is made to vary on a
870 header that is known to contain a wide range of possible values under
871 normal use, for example the <code>User-Agent</code> header. Depending
872 on the popularity of the particular web site thousands or millions of
873 duplicate cache entries could be created for the same URL, crowding
874 out other entries in the cache.</p>
876 <p>In other cases, there may be a need to change the URL of a particular
877 resource on every request, usually by adding a "cachebuster" string to
878 the URL. If this content is declared cacheable by a server for a
879 significant freshness lifetime, these entries can crowd out
880 legitimate entries in a cache. While <module>mod_cache</module>
882 <directive module="mod_cache">CacheIgnoreURLSessionIdentifiers</directive>
883 directive, this directive should be used with care to ensure that
884 downstream proxy or browser caches aren't subjected to the same denial
885 of service issue.</p>