perf-tuning.html revision 777ef9a33550949579958bca5b347f36506f4cdf
<html>
<head>
<title>Apache Performance Notes</title>
</head>
<body bgcolor="#ffffff" text="#000000" link="#0000ff" vlink="#000080" alink="#ff0000">
<h1>Apache Performance Notes</h1>
<p>Author: Dean Gaudet
<h3>Introduction</h3>
<p>Apache is a general webserver, which is designed to be correct first, and
fast second. Even so, it's performance is quite satisfactory. Most
sites have less than 10Mbits of outgoing bandwidth, which Apache can
fill using only a low end Pentium-based webserver. In practice sites
with more bandwidth require more than one machine to fill the bandwidth
due to other constraints (such as CGI or database transaction overhead).
For these reasons the development focus has been mostly on correctness
and configurability.
<p>Unfortunately many folks overlook these facts and cite raw performance
numbers as if they are some indication of the quality of a web server
product. There is a bare minimum performance that is acceptable, beyond
that extra speed only caters to a much smaller segment of the market.
But in order to avoid this hurdle to the acceptance of Apache in some
markets, effort was put into Apache 1.3 to bring performance up to a
point where the difference with other high-end webservers is minimal.
<p>Finally there are the folks who just plain want to see how fast something
can go. The author falls into this category. The rest of this document
is dedicated to these folks who want to squeeze every last bit of
performance out of Apache's current model, and want to understand why
it does some things which slow it down.
<p>Note that this is tailored towards Apache 1.3 on Unix. Some of it applies
to Apache on NT. Apache on NT has not been tuned for performance yet,
in fact it probably performs very poorly because NT performance requires
a different programming model.
<h3>Hardware and Operating System Issues</h3>
<p>The single biggest hardware issue affecting webserver performance
is RAM. A webserver should never ever have to swap, swapping increases
the latency of each request beyond a point that users consider "fast
enough". This causes users to hit stop and reload, further increasing
the load. You can, and should, control the <code>MaxClients</code>
setting so that your server does not spawn so many children it starts
swapping.
<p>Beyond that the rest is mundane: get a fast enough CPU, a fast enough
network card, and fast enough disks, where "fast enough" is something
that needs to be determined by experimentation.
<p>Operating system choice is largely a matter of local concerns. But
HTTP serving completely breaks many of the assumptions built into Unix
kernels up through 1994 and even 1995. Good choices include
recent FreeBSD, and Linux.
<h3>Run-Time Configuration Issues</h3>
<h4>HostnameLookups</h4>
<p>Prior to Apache 1.3, <code>HostnameLookups</code> defaulted to On.
This adds latency
to every request because it requires a DNS lookup to complete before
the request is finished. In Apache 1.3 this setting defaults to Off.
However (1.3 or later), if you use any <code>allow from domain</code> or
<code>deny from domain</code> directives then you will pay for a
double reverse DNS lookup (a reverse, followed by a forward to make sure
that the reverse is not being spoofed). So for the highest performance
avoid using these directives (it's fine to use IP addresses rather than
domain names).
<p>Note that it's possible to scope the directives, such as within
a <code><Location /server-status></code> section. In this
case the DNS lookups are only performed on requests matching the
criteria. Here's an example which disables
lookups except for .html and .cgi files:
<blockquote><pre>
HostnameLookups off
<Files ~ "\.(html|cgi)$>
HostnameLookups on
</Files>
</pre></blockquote>
But even still, if you just need DNS names
in some CGIs you could consider doing the
<code>gethostbyname</code> call in the specific CGIs that need it.
<h4>FollowSymLinks and SymLinksIfOwnerMatch</h4>
<p>Wherever in your URL-space you do not have an
<code>Options FollowSymLinks</code>, or you do have an
<code>Options SymLinksIfOwnerMatch</code> Apache will have to
issue extra system calls to check up on symlinks. One extra call per
filename component. For example, if you had:
<blockquote><pre>
<Directory />
Options SymLinksIfOwnerMatch
</Directory>
</pre></blockquote>
Then Apache will perform <code>lstat(2)</code> on <code>/www</code>,
results of these <code>lstats</code> are never cached,
so they will occur on every single request. If you really desire the
symlinks security checking you can do something like this:
<blockquote><pre>
<Directory />
Options FollowSymLinks
</Directory>
Options -FollowSymLinks +SymLinksIfOwnerMatch
</Directory>
</pre></blockquote>
This at least avoids the extra checks for the <code>DocumentRoot</code>
path. Note that you'll need to add similar sections if you have any
<code>Alias</code> or <code>RewriteRule</code> paths outside of your
document root. For highest performance, and no symlink protection,
set <code>FollowSymLinks</code> everywhere, and never set
<code>SymLinksIfOwnerMatch</code>.
<h4>AllowOverride</h4>
<p>Wherever in your URL-space you allow overrides (typically
<code>.htaccess</code> files) Apache will attempt to open
<code>.htaccess</code> for each filename component. For example,
<blockquote><pre>
<Directory />
AllowOverride all
</Directory>
</pre></blockquote>
Apache will attempt to open <code>/.htaccess</code>,
The solutions are similar to the previous case of <code>Options
FollowSymLinks</code>. For highest performance use
<code>AllowOverride None</code> everywhere in your filesystem.
<h4>Negotiation</h4>
<p>If at all possible, avoid content-negotiation if you're really
interested in every last ounce of performance. In practice the
benefits of negotiation outweigh the performance penalties. There's
one case where you can speed up the server. Instead of using
a wildcard such as:
<blockquote><pre>
DirectoryIndex index
</pre></blockquote>
Use a complete list of options:
<blockquote><pre>
</pre></blockquote>
where you list the most common choice first.
<h4>Process Creation</h4>
<p>Prior to Apache 1.3 the <code>MinSpareServers</code>,
<code>MaxSpareServers</code>, and <code>StartServers</code> settings
all had drastic effects on benchmark results. In particular, Apache
required a "ramp-up" period in order to reach a number of children
sufficient to serve the load being applied. After the initial
spawning of <code>StartServers</code> children, only one child per
second would be created to satisfy the <code>MinSpareServers</code>
setting. So a server being accessed by 100 simultaneous clients,
using the default <code>StartServers</code> of 5 would take on
the order 95 seconds to spawn enough children to handle the load. This
works fine in practice on real-life servers, because they aren't restarted
frequently. But does really poorly on benchmarks which might only run
for ten minutes.
<p>The one-per-second rule was implemented in an effort to avoid
swamping the machine with the startup of new children. If the machine
is busy spawning children it can't service requests. But it has such
a drastic effect on the perceived performance of Apache that it had
to be replaced. As of Apache 1.3,
the code will relax the one-per-second rule. It
will spawn one, wait a second, then spawn two, wait a second, then spawn
four, and it will continue exponentially until it is spawning 32 children
per second. It will stop whenever it satisfies the
<code>MinSpareServers</code> setting.
<p>This appears to be responsive enough that it's
almost unnecessary to twiddle the <code>MinSpareServers</code>,
<code>MaxSpareServers</code> and <code>StartServers</code> knobs. When
more than 4 children are spawned per second, a message will be emitted
to the <code>ErrorLog</code>. If you see a lot of these errors then
consider tuning these settings. Use the <code>mod_status</code> output
as a guide.
<p>Related to process creation is process death induced by the
<code>MaxRequestsPerChild</code> setting. By default this is 30, which
is probably far too low unless your server is using a module such as
<code>mod_perl</code> which causes children to have bloated memory
images. If your server is serving mostly static pages then consider
raising this value to something like 10000. The code is robust enough
that this shouldn't be a problem.
<p>When keep-alives are in use, children will be kept busy
doing nothing waiting for more requests on the already open
connection. The default <code>KeepAliveTimeout</code> of
15 seconds attempts to minimize this effect. The tradeoff
here is between network bandwidth and server resources.
In no event should you raise this above about 60 seconds, as
most of the benefits are lost</a>.
<h3>Compile-Time Configuration Issues</h3>
<h4>mod_status and Rule STATUS=yes</h4>
<p>If you include <code>mod_status</code>
and you also set <code>Rule STATUS=yes</code> when building
Apache, then on every request Apache will perform two calls to
<code>gettimeofday(2)</code> (or <code>times(2)</code> depending
on your operating system), and (pre-1.3) several extra calls to
<code>time(2)</code>. This is all done so that the status report
contains timing indications. For highest performance, set <code>Rule
STATUS=no</code>.
<h4>accept Serialization - multiple sockets</h4>
<p>This discusses a shortcoming in the Unix socket API.
Suppose your
web server uses multiple <code>Listen</code> statements to listen on
either multiple ports or multiple addresses. In order to test each
socket to see if a connection is ready Apache uses <code>select(2)</code>.
<code>select(2)</code> indicates that a socket has <i>none</i> or
<i>at least one</i> connection waiting on it. Apache's model includes
multiple children, and all the idle ones test for new connections at the
same time. A naive implementation looks something like this
(these examples do not match the code, they're contrived for
pedagogical purposes):
<blockquote><pre>
for (;;) {
for (;;) {
fd_set accept_fds;
FD_ZERO (&accept_fds);
for (i = first_socket; i <= last_socket; ++i) {
FD_SET (i, &accept_fds);
}
rc = select (last_socket+1, &accept_fds, NULL, NULL, NULL);
if (rc < 1) continue;
new_connection = -1;
for (i = first_socket; i <= last_socket; ++i) {
if (FD_ISSET (i, &accept_fds)) {
new_connection = accept (i, NULL, NULL);
if (new_connection != -1) break;
}
}
if (new_connection != -1) break;
}
process the new_connection;
}
</pre></blockquote>
But this naive implementation has a serious starvation problem. Recall
that multiple children execute this loop at the same time, and so multiple
children will block at <code>select</code> when they are in between
requests. All those blocked children will awaken and return from
<code>select</code> when a single request appears on any socket
(the number of children which awaken varies depending on the operating
system and timing issues).
They will all then fall down into the loop and try to <code>accept</code>
the connection. But only one will succeed (assuming there's still only
one connection ready), the rest will be <i>blocked</i> in <code>accept</code>.
This effectively locks those children into serving requests from that
one socket and no other sockets, and they'll be stuck there until enough
new requests appear on that socket to wake them all up.
This starvation problem was first documented in
are at least two solutions.
<p>One solution is to make the sockets non-blocking. In this case the
<code>accept</code> won't block the children, and they will be allowed
to continue immediately. But this wastes CPU time. Suppose you have
ten idle children in <code>select</code>, and one connection arrives.
Then nine of those children will wake up, try to <code>accept</code> the
connection, fail, and loop back into <code>select</code>, accomplishing
nothing. Meanwhile none of those children are servicing requests that
occurred on other sockets until they get back up to the <code>select</code>
again. Overall this solution does not seem very fruitful unless you
have as many idle CPUs (in a multiprocessor box) as you have idle children,
not a very likely situation.
<p>Another solution, the one used by Apache, is to serialize entry into
the inner loop. The loop looks like this (differences highlighted):
<blockquote><pre>
for (;;) {
<b>accept_mutex_on ();</b>
for (;;) {
fd_set accept_fds;
FD_ZERO (&accept_fds);
for (i = first_socket; i <= last_socket; ++i) {
FD_SET (i, &accept_fds);
}
rc = select (last_socket+1, &accept_fds, NULL, NULL, NULL);
if (rc < 1) continue;
new_connection = -1;
for (i = first_socket; i <= last_socket; ++i) {
if (FD_ISSET (i, &accept_fds)) {
new_connection = accept (i, NULL, NULL);
if (new_connection != -1) break;
}
}
if (new_connection != -1) break;
}
<b>accept_mutex_off ();</b>
process the new_connection;
}
</pre></blockquote>
<a name="serialize"></a>
The functions <code>accept_mutex_on</code> and <code>accept_mutex_off</code>
implement a mutual exclusion semaphore. Only one child can have the
mutex at any time. There are several choices for implementing these
do not have any locking choice made, on these architectures it is unsafe
to use multiple <code>Listen</code> directives.
<dl>
<dt><code>USE_FLOCK_SERIALIZED_ACCEPT</code>
<dd>This method uses the <code>flock(2)</code> system call to lock a
lock file (located by the <code>LockFile</code> directive).
<dt><code>USE_FCNTL_SERIALIZED_ACCEPT</code>
<dd>This method uses the <code>fcntl(2)</code> system call to lock a
lock file (located by the <code>LockFile</code> directive).
<dt><code>USE_SYSVSEM_SERIALIZED_ACCEPT</code>
<dd>(1.3 or later) This method uses SysV-style semaphores to implement the
mutex. Unfortunately SysV-style semaphores have some bad side-effects.
One is that it's possible Apache will die without cleaning up the semaphore
(see the <code>ipcs(8)</code> man page). The other is that the semaphore
API allows for a denial of service attack by any CGIs running under the
same uid as the webserver (i.e. all CGIs unless you use something
like suexec or cgiwrapper). For these reasons this method is not used
on any architecture except IRIX (where the previous two are prohibitively
expensive on most IRIX boxes).
<dt><code>USE_USLOCK_SERIALIZED_ACCEPT</code>
<dd>(1.3 or later) This method is only available on IRIX, and uses
<code>usconfig(2)</code> to create a mutex. While this method avoids
the hassles of SysV-style semaphores, it is not the default for IRIX.
This is because on single processor IRIX boxes (5.3 or 6.2) the
uslock code is two orders of magnitude slower than the SysV-semaphore
code. On multi-processor IRIX boxes the uslock code is an order of magnitude
faster than the SysV-semaphore code. Kind of a messed up situation.
So if you're using a multiprocessor IRIX box then you should rebuild your
webserver with <code>-DUSE_USLOCK_SERIALIZED_ACCEPT</code> on the
<code>EXTRA_CFLAGS</code>.
<dt><code>USE_PTHREAD_SERIALIZED_ACCEPT</code>
<dd>(1.3 or later) This method uses POSIX mutexes and should work on
any architecture implementing the full POSIX threads specification,
however appears to only work on Solaris (2.5 or later). This is the
default for Solaris 2.5 or later.
</dl>
<p>If your system has another method of serialization which isn't in the
above list then it may be worthwhile adding code for it (and submitting
a patch back to Apache).
<p>Another solution that has been considered but never implemented is
to partially serialize the loop -- that is, let in a certain number
of processes. This would only be of interest on multiprocessor boxes
where it's possible multiple children could run simultaneously, and the
serialization actually doesn't take advantage of the full bandwidth.
This is a possible area of future investigation, but priority remains
low because highly parallel web servers are not the norm.
<p>Ideally you should run servers without multiple <code>Listen</code>
statements if you want the highest performance. But read on.
<h4>accept Serialization - single socket</h4>
<p>The above is fine and dandy for multiple socket servers, but what
about single socket servers? In theory they shouldn't experience
any of these same problems because all children can just block in
<code>accept(2)</code> until a connection arrives, and no starvation
results. In practice this hides almost the same "spinning" behaviour
discussed above in the non-blocking solution. The way that most TCP
stacks are implemented, the kernel actually wakes up all processes blocked
in <code>accept</code> when a single connection arrives. One of those
processes gets the connection and returns to user-space, the rest spin in
the kernel and go back to sleep when they discover there's no connection
for them. This spinning is hidden from the user-land code, but it's
there nonetheless. This can result in the same load-spiking wasteful
behaviour that a non-blocking solution to the multiple sockets case can.
<p>For this reason we have found that many architectures behave more
"nicely" if we serialize even the single socket case. So this is
actually the default in almost all cases. Crude experiments under
the serialization of the single socket case causes less than a 3%
decrease in requests per second over unserialized single-socket.
But unserialized single-socket showed an extra 100ms latency on
each request. This latency is probably a wash on long haul lines,
and only an issue on LANs. If you want to override the single socket
serialization you can define <code>SINGLE_LISTEN_UNSERIALIZED_ACCEPT</code>
and then single-socket servers will not serialize at all.
<h4>Lingering Close</h4>
<p>As discussed in
<a href="ftp://ds.internic.net/internet-drafts/draft-ietf-http-connection-00.txt">draft-ietf-http-connection-00.txt</a> section 8,
in order for an HTTP server to <b>reliably</b> implement the protocol
it needs to shutdown each direction of the communication independently
(recall that a TCP connection is bi-directional, each half is independent
of the other). This fact is often overlooked by other servers, but
is correctly implemented in Apache as of 1.2.
<p>When this feature was added to Apache it caused a flurry of
problems on various versions of Unix because of a shortsightedness.
The TCP specification does not state that the FIN_WAIT_2 state has a
timeout, but it doesn't prohibit it. On systems without the timeout,
Apache 1.2 induces many sockets stuck forever in the FIN_WAIT_2 state.
In many cases this can be avoided by simply upgrading to the latest
never released patches (i.e. SunOS4 -- although folks with a source
license can patch it themselves) we have decided to disable this feature.
<p>There are two ways of accomplishing this. One is the
socket option <code>SO_LINGER</code>. But as fate would have it,
on those stacks with a proper implementation (i.e. Linux 2.0.31) this
method proves to be more expensive (cputime) than the next solution.
<p>For the most part, Apache implements this in a function called
function looks roughly like this:
<blockquote><pre>
void lingering_close (int s)
{
char junk_buffer[2048];
/* shutdown the sending side */
shutdown (s, 1);
signal (SIGALRM, lingering_death);
alarm (30);
for (;;) {
select (s for reading, 2 second timeout);
if (error) break;
if (s is ready for reading) {
read (s, junk_buffer, sizeof (junk_buffer));
/* just toss away whatever is here */
}
}
close (s);
}
</pre></blockquote>
This naturally adds some expense at the end of a connection, but it
prevalent, and all connections are persistent, this expense will be
amortized over more requests. If you want to play with fire and
disable this feature you can define <code>NO_LINGCLOSE</code>, but
persistent connections come into use <code>lingering_close</code>
is an absolute necessity (and
pipelined connections are faster</a>, so you
want to support them).
<h4>Scoreboard File</h4>
<p>Apache's parent and children communicate with each other through
something called the scoreboard. Ideally this should be implemented
in shared memory. For those operating systems that we either have
access to, or have been given detailed ports for, it typically is
implemented using shared memory. The rest default to using an
on-disk file. The on-disk file is not only slow, but it is unreliable
for your architecture and look for either <code>USE_MMAP_SCOREBOARD</code> or
<code>USE_SHMGET_SCOREBOARD</code>. Defining one of those two enables the
supplied shared memory code. If your system has another type of
add the hooks necessary to use it in Apache. (Send us back a patch
too please.)
<p>Historical note: The Linux port of Apache didn't start to use
shared memory until version 1.2 of Apache. This oversight resulted
in really poor and unreliable behaviour of earlier versions of Apache
on Linux.
<h4><code>DYNAMIC_MODULE_LIMIT</code></h4>
<p>If you have no intention of using dynamically loaded modules
(you probably don't if you're reading this and tuning your
server for every last ounce of performance) then you should add
<code>-DDYNAMIC_MODULE_LIMIT=0</code> when building your server.
This will save RAM that's allocated only for supporting dynamically
loaded modules.
<h3>Appendix: Detailed Analysis of a Trace</h3>
Here is a system call trace of Apache 1.3 running on Linux. The run-time
configuration file is essentially the default plus:
<blockquote><pre>
<Directory />
AllowOverride none
Options FollowSymLinks
</Directory>
</pre></blockquote>
The file being requested is a static 6K file of no particular content.
Traces of non-static requests or requests with content negotiation
look wildly different (and quite ugly in some cases). First the
entire trace, then we'll examine details. (This was generated by
the <code>strace</code> program, other similar programs include
<code>truss</code>, <code>ktrace</code>, and <code>par</code>.)
<blockquote><pre>
accept(15, {sin_family=AF_INET, sin_port=htons(22283), sin_addr=inet_addr("127.0.0.1")}, [16]) = 3
flock(18, LOCK_UN) = 0
sigaction(SIGUSR1, {SIG_IGN}, {0x8059954, [], SA_INTERRUPT}) = 0
getsockname(3, {sin_family=AF_INET, sin_port=htons(8080), sin_addr=inet_addr("127.0.0.1")}, [16]) = 0
setsockopt(3, IPPROTO_TCP1, [1], 4) = 0
sigaction(SIGUSR1, {SIG_IGN}, {SIG_IGN}) = 0
time(NULL) = 873959960
gettimeofday({873959960, 404935}, NULL) = 0
mmap(0, 6144, PROT_READ, MAP_PRIVATE, 4, 0) = 0x400ee000
writev(3, [{"HTTP/1.1 200 OK\r\nDate: Thu, 11"..., 245}, {"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 6144}], 2) = 6389
close(4) = 0
time(NULL) = 873959960
gettimeofday({873959960, 417742}, NULL) = 0
times({tms_utime=5, tms_stime=0, tms_cutime=0, tms_cstime=0}) = 446747
shutdown(3, 1 /* send */) = 0
oldselect(4, [3], NULL, [3], {2, 0}) = 1 (in [3], left {2, 0})
read(3, "", 2048) = 0
close(3) = 0
sigaction(SIGUSR1, {0x8059954, [], SA_INTERRUPT}, {SIG_IGN}) = 0
munmap(0x400ee000, 6144) = 0
flock(18, LOCK_EX) = 0
</pre></blockquote>
<p>Notice the accept serialization:
<blockquote><pre>
flock(18, LOCK_UN) = 0
...
flock(18, LOCK_EX) = 0
</pre></blockquote>
These two calls can be removed by defining
<code>SINGLE_LISTEN_UNSERIALIZED_ACCEPT</code> as described earlier.
<p>Notice the <code>SIGUSR1</code> manipulation:
<blockquote><pre>
sigaction(SIGUSR1, {SIG_IGN}, {0x8059954, [], SA_INTERRUPT}) = 0
...
sigaction(SIGUSR1, {SIG_IGN}, {SIG_IGN}) = 0
...
sigaction(SIGUSR1, {0x8059954, [], SA_INTERRUPT}, {SIG_IGN}) = 0
</pre></blockquote>
This is caused by the implementation of graceful restarts. When the
parent receives a <code>SIGUSR1</code> it sends a <code>SIGUSR1</code>
to all of its children (and it also increments a "generation counter"
in shared memory). Any children that are idle (between connections)
will immediately die
off when they receive the signal. Any children that are in keep-alive
connections, but are in between requests will die off immediately. But
any children that have a connection and are still waiting for the first
request will not die off immediately.
<p>To see why this is necessary, consider how a browser reacts to a closed
connection. If the connection was a keep-alive connection and the request
being serviced was not the first request then the browser will quietly
reissue the request on a new connection. It has to do this because the
server is always free to close a keep-alive connection in between requests
(i.e. due to a timeout or because of a maximum number of requests).
But, if the connection is closed before the first response has been
received the typical browser will display a "document contains no data"
dialogue (or a broken image icon). This is done on the assumption that
the server is broken in some way (or maybe too overloaded to respond
at all). So Apache tries to avoid ever deliberately closing the connection
before it has sent a single response. This is the cause of those
<code>SIGUSR1</code> manipulations.
<p>Note that it is theoretically possible to eliminate all three of
these calls. But in rough tests the gain proved to be almost unnoticeable.
<p>In order to implement virtual hosts, Apache needs to know the
local socket address used to accept the connection:
<blockquote><pre>
getsockname(3, {sin_family=AF_INET, sin_port=htons(8080), sin_addr=inet_addr("127.0.0.1")}, [16]) = 0
</pre></blockquote>
It is possible to eliminate this call in many situations (such as when
there are no virtual hosts, or when <code>Listen</code> directives are
used which do not have wildcard addresses). But no effort has yet been
made to do these optimizations.
<p>Apache turns off the Nagle algorithm:
<blockquote><pre>
setsockopt(3, IPPROTO_TCP1, [1], 4) = 0
</pre></blockquote>
because of problems described in
paper by John Heidemann</a>.
<p>Notice the two <code>time</code> calls:
<blockquote><pre>
time(NULL) = 873959960
...
time(NULL) = 873959960
</pre></blockquote>
One of these occurs at the beginning of the request, and the other occurs
as a result of writing the log. At least one of these is required to
properly implement the HTTP protocol. The second occurs because the
Common Log Format dictates that the log record include a timestamp of the
end of the request. A custom logging module could eliminate one of the
calls.
<p>As described earlier, <code>Rule STATUS=yes</code> causes two
<code>gettimeofday</code> calls and a call to <code>times</code>:
<blockquote><pre>
gettimeofday({873959960, 404935}, NULL) = 0
...
gettimeofday({873959960, 417742}, NULL) = 0
times({tms_utime=5, tms_stime=0, tms_cutime=0, tms_cstime=0}) = 446747
</pre></blockquote>
These can be removed by either removing <code>mod_status</code> or
setting <code>Rule STATUS=no</code>.
<p>It might seem odd to call <code>stat</code>:
<blockquote><pre>
</pre></blockquote>
This is part of the algorithm which calculates the
<code>PATH_INFO</code> for use by CGIs. In fact if the request had been
two calls to <code>stat</code>. The first for
which does not exist, and the second for
Regardless, at least one <code>stat</code> call is necessary when
serving static files because the file size and modification times are
used to generate HTTP headers (such as <code>Content-Length</code>,
<code>Last-Modified</code>) and implement protocol features (such
as <code>If-Modified-Since</code>). A somewhat more clever server
could avoid the <code>stat</code> when serving non-static files,
however doing so in Apache is very difficult given the modular structure.
<p>All static files are served using <code>mmap</code>:
<blockquote><pre>
mmap(0, 6144, PROT_READ, MAP_PRIVATE, 4, 0) = 0x400ee000
...
munmap(0x400ee000, 6144) = 0
</pre></blockquote>
On some architectures it's slower to <code>mmap</code> small
files than it is to simply <code>read</code> them. The define
<code>MMAP_THRESHOLD</code> can be set to the minimum
size required before using <code>mmap</code>. By default
it's set to 0 (except on SunOS4 where experimentation has
shown 8192 to be a better value). Using a tool such as <a
href="http://reality.sgi.com/lm_engr/lmbench/lmbench.html">lmbench</a> you
can determine the optimal setting for your environment.
<p>You may also wish to experiment with <code>MMAP_SEGMENT_SIZE</code>
(default 32768) which determines the maximum number of bytes that
will be written at a time from mmap()d files. Apache only resets the
client's <code>Timeout</code> in between write()s. So setting this
large may lock out low bandwidth clients unless you also increase the
<code>Timeout</code>.
<p>It may even be the case that <code>mmap</code> isn't
used on your architecture, if so then defining <code>USE_MMAP_FILES</code>
might work (if it works then report back to us).
<p>Apache does its best to avoid copying bytes around in memory. The
first write of any request typically is turned into a <code>writev</code>
which combines both the headers and the first hunk of data:
<blockquote><pre>
writev(3, [{"HTTP/1.1 200 OK\r\nDate: Thu, 11"..., 245}, {"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 6144}], 2) = 6389
</pre></blockquote>
element <code>writev</code>s. The goal is to push the byte copying
into the kernel, where it typically has to happen anyhow (to assemble
network packets). On testing, various Unixes (BSDI 2.x, Solaris 2.5,
Linux 2.0.31+) properly combine the elements into network packets.
Pre-2.0.31 Linux will not combine, and will create a packet for
each element, so upgrading is a good idea. Defining <code>NO_WRITEV</code>
will disable this combining, but result in very poor chunked encoding
performance.
<p>The log write:
<blockquote><pre>
</pre></blockquote>
can be deferred by defining <code>BUFFERED_LOGS</code>. In this case
up to <code>PIPE_BUF</code> bytes (a POSIX defined constant) of log entries
are buffered before writing. At no time does it split a log entry
across a <code>PIPE_BUF</code> boundary because those writes may not
be atomic. (i.e. entries from multiple children could become mixed together).
The code does it best to flush this buffer when a child dies.
<p>The lingering close code causes four system calls:
<blockquote><pre>
shutdown(3, 1 /* send */) = 0
oldselect(4, [3], NULL, [3], {2, 0}) = 1 (in [3], left {2, 0})
read(3, "", 2048) = 0
close(3) = 0
</pre></blockquote>
which were described earlier.
<p>Let's apply some of these optimizations:
<code>-DSINGLE_LISTEN_UNSERIALIZED_ACCEPT -DBUFFERED_LOGS</code> and
<code>Rule STATUS=no</code>. Here's the final trace:
<blockquote><pre>
accept(15, {sin_family=AF_INET, sin_port=htons(22286), sin_addr=inet_addr("127.0.0.1")}, [16]) = 3
sigaction(SIGUSR1, {SIG_IGN}, {0x8058c98, [], SA_INTERRUPT}) = 0
getsockname(3, {sin_family=AF_INET, sin_port=htons(8080), sin_addr=inet_addr("127.0.0.1")}, [16]) = 0
setsockopt(3, IPPROTO_TCP1, [1], 4) = 0
sigaction(SIGUSR1, {SIG_IGN}, {SIG_IGN}) = 0
time(NULL) = 873961916
mmap(0, 6144, PROT_READ, MAP_PRIVATE, 4, 0) = 0x400e3000
writev(3, [{"HTTP/1.1 200 OK\r\nDate: Thu, 11"..., 245}, {"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 6144}], 2) = 6389
close(4) = 0
time(NULL) = 873961916
shutdown(3, 1 /* send */) = 0
oldselect(4, [3], NULL, [3], {2, 0}) = 1 (in [3], left {2, 0})
read(3, "", 2048) = 0
close(3) = 0
sigaction(SIGUSR1, {0x8058c98, [], SA_INTERRUPT}, {SIG_IGN}) = 0
munmap(0x400e3000, 6144) = 0
</pre></blockquote>
That's 19 system calls, of which 4 remain relatively easy to remove,
but don't seem worth the effort.
<h3>Appendix: The Pre-Forking Model</h3>
<p>Apache (on Unix) is a <i>pre-forking</i> model server. The
<i>parent</i> process is responsible only for forking <i>child</i>
processes, it does not serve any requests or service any network
sockets. The child processes actually process connections, they serve
multiple connections (one at a time) before dying.
The parent spawns new or kills off old
children in response to changes in the load on the server (it does so
by monitoring a scoreboard which the children keep up to date).
<p>This model for servers offers a robustness that other models do
not. In particular, the parent code is very simple, and with a high
degree of confidence the parent will continue to do its job without
error. The children are complex, and when you add in third party
code via modules, you risk segmentation faults and other forms of
corruption. Even should such a thing happen, it only affects one
connection and the server continues serving requests. The parent
quickly replaces the dead child.
<p>Pre-forking is also very portable across dialects of Unix.
Historically this has been an important goal for Apache, and it continues
to remain so.
<p>The pre-forking model comes under criticism for various
performance aspects. Of particular concern are the overhead
of forking a process, the overhead of context switches between
processes, and the memory overhead of having multiple processes.
Furthermore it does not offer as many opportunities for data-caching
between requests (such as a pool of <code>mmapped</code> files).
Various other models exist and extensive analysis can be found in the
<a href="http://www.cs.wustl.edu/~jxh/research/research.html"> papers
of the JAWS project</a>. In practice all of these costs vary drastically
depending on the operating system.
<p>Apache's core code is already multithread aware, and Apache version
1.3 is multithreaded on NT. There have been at least two other experimental
implementations of threaded Apache (one using the 1.3 code base on DCE,
and one using a custom user-level threads package and the 1.0 code base,
neither are available publically). Part of our redesign for version 2.0
of Apache will include abstractions of the server model so that we
can continue to support the pre-forking model, and also support various
threaded models.
</body>
</html>