httpcomponents-client/src/docbkx/caching.xml

254 lines
11 KiB
XML

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE preface PUBLIC "-//OASIS//DTD DocBook XML V4.4//EN"
"http://www.oasis-open.org/docbook/xml/4.4/docbookx.dtd">
<!--
====================================================================
Licensed to the Apache Software Foundation (ASF) under one
or more contributor license agreements. See the NOTICE file
distributed with this work for additional information
regarding copyright ownership. The ASF licenses this file
to you under the Apache License, Version 2.0 (the
"License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
KIND, either express or implied. See the License for the
specific language governing permissions and limitations
under the License.
====================================================================
-->
<chapter id="caching">
<title>HTTP Caching</title>
<section id="generalconcepts">
<title>General Concepts</title>
<para>HttpClient Cache provides an HTTP/1.1-compliant caching layer to be
used with HttpClient--the Java equivalent of a browser cache. The
implementation follows the Chain of Responsibility design pattern, where the
caching HttpClient implementation can serve a drop-in replacement for
the default non-caching HttpClient implementation; requests that can be
satisfied entirely from the cache will not result in actual origin requests.
Stale cache entries are automatically validated with the origin where possible,
using conditional GETs and the If-Modified-Since and/or If-None-Match request
headers.
</para>
<para>
HTTP/1.1 caching in general is designed to be <emphasis>semantically
transparent</emphasis>; that is, a cache should not change the meaning of
the request-response exchange between client and server. As such, it should
be safe to drop a caching HttpClient into an existing compliant client-server
relationship. Although the caching module is part of the client from an
HTTP protocol point of view, the implementation aims to be compatible with
the requirements placed on a transparent caching proxy.
</para>
<para>Finally, caching HttpClient includes support the Cache-Control
extensions specified by RFC 5861 (stale-if-error and stale-while-revalidate).
</para>
<para>When caching HttpClient executes a request, it goes through the
following flow:</para>
<orderedlist>
<listitem>
<para>Check the request for basic compliance with the HTTP 1.1
protocol and attempt to correct the request.</para>
</listitem>
<listitem>
<para>Flush any cache entries which would be invalidated by this
request.</para>
</listitem>
<listitem>
<para>Determine if the current request would be servable from cache.
If not, directly pass through the request to the origin server and
return the response, after caching it if appropriate.</para>
</listitem>
<listitem>
<para>If it was a a cache-servable request, it will attempt to read it
from the cache. If it is not in the cache, call the origin server and
cache the response, if appropriate.</para>
</listitem>
<listitem>
<para>If the cached response is suitable to be served as a response,
construct a BasicHttpResponse containing a ByteArrayEntity and return
it. Otherwise, attempt to revalidate the cache entry against the
origin server.</para>
</listitem>
<listitem>
<para>In the case of a cached response which cannot be revalidated,
call the origin server and cache the response, if appropriate.</para>
</listitem>
</orderedlist>
<para>When caching HttpClient receives a response, it goes through the
following flow:</para>
<orderedlist>
<listitem>
<para>Examining the response for protocol compliance</para>
</listitem>
<listitem>
<para>Determine whether the response is cacheable</para>
</listitem>
<listitem>
<para>If it is cacheable, attempt to read up to the maximum size
allowed in the configuration and store it in the cache.</para>
</listitem>
<listitem>
<para>If the response is too large for the cache, reconstruct the
partially consumed response and return it directly without caching
it.</para>
</listitem>
</orderedlist>
<para>It is important to note that caching HttpClient is not, itself,
a different implementation of HttpClient, but that it works by inserting
itself as an additonal processing component to the request execution
pipeline.</para>
</section>
<section id="rfc2616compliance">
<title>RFC-2616 Compliance</title>
<para>We believe HttpClient Cache is <emphasis>unconditionally
compliant</emphasis> with <ulink
url="http://www.ietf.org/rfc/rfc2616.txt">RFC-2616</ulink>. That is,
wherever the specification indicates MUST, MUST NOT, SHOULD, or SHOULD NOT
for HTTP caches, the caching layer attempts to behave in a way that satisfies
those requirements. This means the caching module won't produce incorrect
behavior when you drop it in. </para>
</section>
<section>
<title>Example Usage</title>
<para>This is a simple example of how to set up a basic caching HttpClient.
As configured, it will store a maximum of 1000 cached objects, each of
which may have a maximum body size of 8192 bytes. The numbers selected
here are for example only and not intended to be prescriptive or
considered as recommendations.</para>
<programlisting><![CDATA[
CacheConfig cacheConfig = CacheConfig.custom()
.setMaxCacheEntries(1000)
.setMaxObjectSize(8192)
.build();
RequestConfig requestConfig = RequestConfig.custom()
.setConnectTimeout(30000)
.setSocketTimeout(30000)
.build();
CloseableHttpClient cachingClient = CachingHttpClients.custom()
.setCacheConfig(cacheConfig)
.setDefaultRequestConfig(requestConfig)
.build();
HttpCacheContext context = HttpCacheContext.create();
HttpGet httpget = new HttpGet("http://www.mydomain.com/content/");
CloseableHttpResponse response = cachingClient.execute(httpget, context);
try {
CacheResponseStatus responseStatus = context.getCacheResponseStatus();
switch (responseStatus) {
case CACHE_HIT:
System.out.println("A response was generated from the cache with " +
"no requests sent upstream");
break;
case CACHE_MODULE_RESPONSE:
System.out.println("The response was generated directly by the " +
"caching module");
break;
case CACHE_MISS:
System.out.println("The response came from an upstream server");
break;
case VALIDATED:
System.out.println("The response was generated from the cache " +
"after validating the entry with the origin server");
break;
}
} finally {
response.close();
}
]]>
</programlisting>
</section>
<section id="configuration">
<title>Configuration</title>
<para>The caching HttpClient inherits all configuration options and parameters
of the default non-caching implementation (this includes setting options like
timeouts and connection pool sizes). For caching-specific configuration, you can
provide a <classname>CacheConfig</classname> instance to customize behavior
across the following areas:</para>
<para><emphasis>Cache size.</emphasis> If the backend storage supports these limits,
you can specify the maximum number of cache entries as well as the maximum cacheable
response body size.</para>
<para><emphasis>Public/private caching.</emphasis> By default, the caching module
considers itself to be a shared (public) cache, and will not, for example, cache
responses to requests with Authorization headers or responses marked with
"Cache-Control: private". If, however, the cache is only going to be used by one
logical "user" (behaving similarly to a browser cache), then you will want to turn
off the shared cache setting.</para>
<para><emphasis>Heuristic caching.</emphasis>Per RFC2616, a cache MAY cache
certain cache entries even if no explicit cache control headers are set by the
origin. This behavior is off by default, but you may want to turn this on if you
are working with an origin that doesn't set proper headers but where you still
want to cache the responses. You will want to enable heuristic caching, then
specify either a default freshness lifetime and/or a fraction of the time since
the resource was last modified. See Sections 13.2.2 and 13.2.4 of the HTTP/1.1
RFC for more details on heuristic caching.</para>
<para><emphasis>Background validation.</emphasis> The cache module supports the
stale-while-revalidate directive of RFC5861, which allows certain cache entry
revalidations to happen in the background. You may want to tweak the settings
for the minimum and maximum number of background worker threads, as well as the
maximum time they can be idle before being reclaimed. You can also control the
size of the queue used for revalidations when there aren't enough workers to
keep up with demand.</para>
</section>
<section id="storage">
<title>Storage Backends</title>
<para>The default implementation of caching HttpClient stores cache entries and
cached response bodies in memory in the JVM of your application. While this
offers high performance, it may not be appropriate for your application due to
the limitation on size or because the cache entries are ephemeral and don't
survive an application restart. The current release includes support for storing
cache entries using EhCache and memcached implementations, which allow for
spilling cache entries to disk or storing them in an external process.</para>
<para>If none of those options are suitable for your application, it is
possible to provide your own storage backend by implementing the HttpCacheStorage
interface and then supplying that to caching HttpClient at construction time. In
this case, the cache entries will be stored using your scheme but you will get to
reuse all of the logic surrounding HTTP/1.1 compliance and cache handling.
Generally speaking, it should be possible to create an HttpCacheStorage
implementation out of anything that supports a key/value store (similar to the
Java Map interface) with the ability to apply atomic updates.</para>
<para>Finally, with some extra efforts it's entirely possible to set up
a multi-tier caching hierarchy; for example, wrapping an in-memory caching
HttpClient around one that stores cache entries on disk or remotely in memcached,
following a pattern similar to virtual memory, L1/L2 processor caches, etc.
</para>
</section>
</chapter>