1 <?xml version="1.0" encoding="ISO-8859-1"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en"><head><!--
4 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
5 This file is generated from xml source: DO NOT EDIT
6 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
8 <title>mod_unique_id - Apache HTTP Server</title>
9 <link href="../style/css/manual.css" rel="stylesheet" media="all" type="text/css" title="Main stylesheet" />
10 <link href="../style/css/manual-loose-100pc.css" rel="alternate stylesheet" media="all" type="text/css" title="No Sidebar - Default font size" />
11 <link href="../style/css/manual-print.css" rel="stylesheet" media="print" type="text/css" />
12 <link href="../images/favicon.ico" rel="shortcut icon" /></head>
14 <div id="page-header">
15 <p class="menu"><a href="../mod/">Modules</a> | <a href="../mod/directives.html">Directives</a> | <a href="../faq/">FAQ</a> | <a href="../glossary.html">Glossary</a> | <a href="../sitemap.html">Sitemap</a></p>
16 <p class="apache">Apache HTTP Server Version 2.5</p>
17 <img alt="" src="../images/feather.gif" /></div>
18 <div class="up"><a href="./"><img title="<-" alt="<-" src="../images/left.gif" /></a></div>
20 <a href="http://www.apache.org/">Apache</a> > <a href="http://httpd.apache.org/">HTTP Server</a> > <a href="http://httpd.apache.org/docs/">Documentation</a> > <a href="../">Version 2.5</a> > <a href="./">Modules</a></div>
21 <div id="page-content">
22 <div id="preamble"><h1>Apache Module mod_unique_id</h1>
24 <p><span>Available Languages: </span><a href="../en/mod/mod_unique_id.html" title="English"> en </a> |
25 <a href="../fr/mod/mod_unique_id.html" hreflang="fr" rel="alternate" title="Français"> fr </a> |
26 <a href="../ja/mod/mod_unique_id.html" hreflang="ja" rel="alternate" title="Japanese"> ja </a> |
27 <a href="../ko/mod/mod_unique_id.html" hreflang="ko" rel="alternate" title="Korean"> ko </a></p>
29 <table class="module"><tr><th><a href="module-dict.html#Description">Description:</a></th><td>Provides an environment variable with a unique
30 identifier for each request</td></tr>
31 <tr><th><a href="module-dict.html#Status">Status:</a></th><td>Extension</td></tr>
32 <tr><th><a href="module-dict.html#ModuleIdentifier">Module Identifier:</a></th><td>unique_id_module</td></tr>
33 <tr><th><a href="module-dict.html#SourceFile">Source File:</a></th><td>mod_unique_id.c</td></tr></table>
37 <p>This module provides a magic token for each request which is
38 guaranteed to be unique across "all" requests under very
39 specific conditions. The unique identifier is even unique
40 across multiple machines in a properly configured cluster of
41 machines. The environment variable <code>UNIQUE_ID</code> is
42 set to the identifier for each request. Unique identifiers are
43 useful for various reasons which are beyond the scope of this
46 <div id="quickview"><h3 class="directives">Directives</h3>
47 <p>This module provides no
51 <li><img alt="" src="../images/down.gif" /> <a href="#theory">Theory</a></li>
53 <div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
55 <h2><a name="theory" id="theory">Theory</a></h2>
58 <p>First a brief recap of how the Apache server works on Unix
59 machines. This feature currently isn't supported on Windows NT.
60 On Unix machines, Apache creates several children, the children
61 process requests one at a time. Each child can serve multiple
62 requests in its lifetime. For the purpose of this discussion,
63 the children don't share any data with each other. We'll refer
64 to the children as <dfn>httpd processes</dfn>.</p>
66 <p>Your website has one or more machines under your
67 administrative control, together we'll call them a cluster of
68 machines. Each machine can possibly run multiple instances of
69 Apache. All of these collectively are considered "the
70 universe", and with certain assumptions we'll show that in this
71 universe we can generate unique identifiers for each request,
72 without extensive communication between machines in the
75 <p>The machines in your cluster should satisfy these
76 requirements. (Even if you have only one machine you should
77 synchronize its clock with NTP.)</p>
80 <li>The machines' times are synchronized via NTP or other
81 network time protocol.</li>
83 <li>The machines' hostnames all differ, such that the module
84 can do a hostname lookup on the hostname and receive a
85 different IP address for each machine in the cluster.</li>
88 <p>As far as operating system assumptions go, we assume that
89 pids (process ids) fit in 32-bits. If the operating system uses
90 more than 32-bits for a pid, the fix is trivial but must be
91 performed in the code.</p>
93 <p>Given those assumptions, at a single point in time we can
94 identify any httpd process on any machine in the cluster from
95 all other httpd processes. The machine's IP address and the pid
96 of the httpd process are sufficient to do this. A httpd process
97 can handle multiple requests simultaneously if you use a
98 multi-threaded MPM. In order to identify threads, we use a thread
99 index Apache httpd uses internally. So in order to
100 generate unique identifiers for requests we need only
101 distinguish between different points in time.</p>
103 <p>To distinguish time we will use a Unix timestamp (seconds
104 since January 1, 1970 UTC), and a 16-bit counter. The timestamp
105 has only one second granularity, so the counter is used to
106 represent up to 65536 values during a single second. The
107 quadruple <em>( ip_addr, pid, time_stamp, counter )</em> is
108 sufficient to enumerate 65536 requests per second per httpd
109 process. There are issues however with pid reuse over time, and
110 the counter is used to alleviate this issue.</p>
112 <p>When an httpd child is created, the counter is initialized
113 with ( current microseconds divided by 10 ) modulo 65536 (this
114 formula was chosen to eliminate some variance problems with the
115 low order bits of the microsecond timers on some systems). When
116 a unique identifier is generated, the time stamp used is the
117 time the request arrived at the web server. The counter is
118 incremented every time an identifier is generated (and allowed
121 <p>The kernel generates a pid for each process as it forks the
122 process, and pids are allowed to roll over (they're 16-bits on
123 many Unixes, but newer systems have expanded to 32-bits). So
124 over time the same pid will be reused. However unless it is
125 reused within the same second, it does not destroy the
126 uniqueness of our quadruple. That is, we assume the system does
127 not spawn 65536 processes in a one second interval (it may even
128 be 32768 processes on some Unixes, but even this isn't likely
131 <p>Suppose that time repeats itself for some reason. That is,
132 suppose that the system's clock is screwed up and it revisits a
133 past time (or it is too far forward, is reset correctly, and
134 then revisits the future time). In this case we can easily show
135 that we can get pid and time stamp reuse. The choice of
136 initializer for the counter is intended to help defeat this.
137 Note that we really want a random number to initialize the
138 counter, but there aren't any readily available numbers on most
139 systems (<em>i.e.</em>, you can't use rand() because you need
140 to seed the generator, and can't seed it with the time because
141 time, at least at one second resolution, has repeated itself).
142 This is not a perfect defense.</p>
144 <p>How good a defense is it? Suppose that one of your machines
145 serves at most 500 requests per second (which is a very
146 reasonable upper bound at this writing, because systems
147 generally do more than just shovel out static files). To do
148 that it will require a number of children which depends on how
149 many concurrent clients you have. But we'll be pessimistic and
150 suppose that a single child is able to serve 500 requests per
151 second. There are 1000 possible starting counter values such
152 that two sequences of 500 requests overlap. So there is a 1.5%
153 chance that if time (at one second resolution) repeats itself
154 this child will repeat a counter value, and uniqueness will be
155 broken. This was a very pessimistic example, and with real
156 world values it's even less likely to occur. If your system is
157 such that it's still likely to occur, then perhaps you should
158 make the counter 32 bits (by editing the code).</p>
160 <p>You may be concerned about the clock being "set back" during
161 summer daylight savings. However this isn't an issue because
162 the times used here are UTC, which "always" go forward. Note
163 that x86 based Unixes may need proper configuration for this to
164 be true -- they should be configured to assume that the
165 motherboard clock is on UTC and compensate appropriately. But
166 even still, if you're running NTP then your UTC time will be
167 correct very shortly after reboot.</p>
170 <p>The <code>UNIQUE_ID</code> environment variable is
171 constructed by encoding the 144-bit (32-bit IP address, 32 bit
172 pid, 32 bit time stamp, 16 bit counter, 32 bit thread index)
174 alphabet <code>[A-Za-z0-9@-]</code> in a manner similar to MIME
175 base64 encoding, producing 24 characters. The MIME base64
176 alphabet is actually <code>[A-Za-z0-9+/]</code> however
177 <code>+</code> and <code>/</code> need to be specially encoded
178 in URLs, which makes them less desirable. All values are
179 encoded in network byte ordering so that the encoding is
180 comparable across architectures of different byte ordering. The
181 actual ordering of the encoding is: time stamp, IP address,
182 pid, counter. This ordering has a purpose, but it should be
183 emphasized that applications should not dissect the encoding.
184 Applications should treat the entire encoded
185 <code>UNIQUE_ID</code> as an opaque token, which can be
186 compared against other <code>UNIQUE_ID</code>s for equality
189 <p>The ordering was chosen such that it's possible to change
190 the encoding in the future without worrying about collision
191 with an existing database of <code>UNIQUE_ID</code>s. The new
192 encodings should also keep the time stamp as the first element,
193 and can otherwise use the same alphabet and bit length. Since
194 the time stamps are essentially an increasing sequence, it's
195 sufficient to have a <em>flag second</em> in which all machines
196 in the cluster stop serving and request, and stop using the old
197 encoding format. Afterwards they can resume requests and begin
198 issuing the new encodings.</p>
200 <p>This we believe is a relatively portable solution to this
201 problem. The identifiers
202 generated have essentially an infinite life-time because future
203 identifiers can be made longer as required. Essentially no
204 communication is required between machines in the cluster (only
205 NTP synchronization is required, which is low overhead), and no
206 communication between httpd processes is required (the
207 communication is implicit in the pid value assigned by the
208 kernel). In very specific situations the identifier can be
209 shortened, but more information needs to be assumed (for
210 example the 32-bit IP address is overkill for any site, but
211 there is no portable shorter replacement for it). </p>
214 <div class="bottomlang">
215 <p><span>Available Languages: </span><a href="../en/mod/mod_unique_id.html" title="English"> en </a> |
216 <a href="../fr/mod/mod_unique_id.html" hreflang="fr" rel="alternate" title="Français"> fr </a> |
217 <a href="../ja/mod/mod_unique_id.html" hreflang="ja" rel="alternate" title="Japanese"> ja </a> |
218 <a href="../ko/mod/mod_unique_id.html" hreflang="ko" rel="alternate" title="Korean"> ko </a></p>
219 </div><div id="footer">
220 <p class="apache">Copyright 2011 The Apache Software Foundation.<br />Licensed under the <a href="http://www.apache.org/licenses/LICENSE-2.0">Apache License, Version 2.0</a>.</p>
221 <p class="menu"><a href="../mod/">Modules</a> | <a href="../mod/directives.html">Directives</a> | <a href="../faq/">FAQ</a> | <a href="../glossary.html">Glossary</a> | <a href="../sitemap.html">Sitemap</a></p></div>