Compile-Time Configuration Issues

Choosing an MPM

Apache 2.x supports pluggable concurrency models, called Multi-Processing Modules (MPMs). When building Apache, you must choose an MPM to use. There are platform-specific MPMs for some platforms: beos, mpm_netware, mpmt_os2, and mpm_winnt. For general Unix-type systems, there are several MPMs from which to choose. The choice of MPM can affect the speed and scalability of the httpd:

• The worker MPM uses multiple child processes with many threads each. Each thread handles one connection at a time. Worker generally is a good choice for high-traffic servers because it has a smaller memory footprint than the prefork MPM.
• The prefork MPM uses multiple child processes with one thread each. Each process handles one connection at a time. On many systems, prefork is comparable in speed to worker, but it uses more memory. Prefork's threadless design has advantages over worker in some situations: it can be used with non-thread-safe third-party modules, and it is easier to debug on platforms with poor thread debugging support.

For more information on these and other MPMs, please see the MPM documentation.

Modules

Since memory usage is such an important consideration in performance, you should attempt to eliminate modules that you are not actually using. If you have built the modules as DSOs, eliminating modules is a simple matter of commenting out the associated LoadModule directive for that module. This allows you to experiment with removing modules and seeing if your site still functions in their absense.

If, on the other hand, you have modules statically linked into your Apache binary, you will need to recompile Apache in order to remove unwanted modules.

An associated question that arises here is, of course, what modules you need, and which ones you don't. The answer here will, of course, vary from one web site to another. However, the minimal list of modules which you can get by with tends to include mod_mime, mod_dir, and mod_log_config. mod_log_config is, of course, optional, as you can run a web site without log files. This is, however, not recommended.

Atomic Operations

Some modules, such as mod_cache and recent development builds of the worker MPM, use APR's atomic API. This API provides atomic operations that can be used for lightweight thread synchronization.

By default, APR implements these operations using the most efficient mechanism available on each target OS/CPU platform. Many modern CPUs, for example, have an instruction that does an atomic compare-and-swap (CAS) operation in hardware. On some platforms, however, APR defaults to a slower, mutex-based implementation of the atomic API in order to ensure compatibility with older CPU models that lack such instructions. If you are building Apache for one of these platforms, and you plan to run only on newer CPUs, you can select a faster atomic implementation at build time by configuring Apache with the --enable-nonportable-atomics option:

The --enable-nonportable-atomics option is relevant for the following platforms:

• Solaris on SPARC
  By default, APR uses mutex-based atomics on Solaris/SPARC. If you configure with --enable-nonportable-atomics, however, APR generates code that uses a SPARC v8plus opcode for fast hardware compare-and-swap. If you configure Apache with this option, the atomic operations will be more efficient (allowing for lower CPU utilization and higher concurrency), but the resulting executable will run only on UltraSPARC chips.
• Linux on x86
  By default, APR uses mutex-based atomics on Linux. If you configure with --enable-nonportable-atomics, however, APR generates code that uses a 486 opcode for fast hardware compare-and-swap. This will result in more efficient atomic operations, but the resulting executable will run only on 486 and later chips (and not on 386).

mod_status and ExtendedStatus On

If you include mod_status and you also set ExtendedStatus On when building and running Apache, then on every request Apache will perform two calls to gettimeofday(2) (or times(2) depending on your operating system), and (pre-1.3) several extra calls to time(2). This is all done so that the status report contains timing indications. For highest performance, set ExtendedStatus off (which is the default).

accept Serialization - Multiple Sockets

This discusses a shortcoming in the Unix socket API. Suppose your web server uses multiple Listen 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 select(2). select(2) indicates that a socket has zero or at least one 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):

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 select when they are in between requests. All those blocked children will awaken and return from select 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 accept the connection. But only one will succeed (assuming there's still only one connection ready). The rest will be blocked in accept. 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 PR#467. There are at least two solutions.

One solution is to make the sockets non-blocking. In this case the accept 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 select, and one connection arrives. Then nine of those children will wake up, try to accept the connection, fail, and loop back into select, accomplishing nothing. Meanwhile none of those children are servicing requests that occurred on other sockets until they get back up to the select 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).

Another solution, the one used by Apache, is to serialize entry into the inner loop. The loop looks like this (differences highlighted):

The functions accept_mutex_on and accept_mutex_off implement a mutual exclusion semaphore. Only one child can have the mutex at any time. There are several choices for implementing these mutexes. The choice is defined in src/conf.h (pre-1.3) or src/include/ap_config.h (1.3 or later). Some architectures do not have any locking choice made, on these architectures it is unsafe to use multiple Listen directives.

The directive AcceptMutex can be used to change the selected mutex implementation at run-time.

AcceptMutex flock

This method uses the flock(2) system call to lock a lock file (located by the LockFile directive).

AcceptMutex fcntl

This method uses the fcntl(2) system call to lock a lock file (located by the LockFile directive).

AcceptMutex sysvsem

(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 ipcs(8) 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).

AcceptMutex pthread

(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), and even then only in certain configurations. If you experiment with this you should watch out for your server hanging and not responding. Static content only servers may work just fine.

AcceptMutex posixsem

(2.0 or later) This method uses POSIX semaphores. The semaphore ownership is not recovered if a thread in the process holding the mutex segfaults, resulting in a hang of the web server.

If your system has another method of serialization which isn't in the above list then it may be worthwhile adding code for it to APR.

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 that 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.

Ideally you should run servers without multiple Listen statements if you want the highest performance. But read on.

accept Serialization - Single Socket

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 accept(2) until a connection arrives, and no starvation results. In practice this hides almost the same "spinning" behavior discussed above in the non-blocking solution. The way that most TCP stacks are implemented, the kernel actually wakes up all processes blocked in accept 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 behavior that a non-blocking solution to the multiple sockets case can.

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 Linux (2.0.30 on a dual Pentium pro 166 w/128Mb RAM) have shown that 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 SINGLE_LISTEN_UNSERIALIZED_ACCEPT, and then single-socket servers will not serialize at all.

Lingering Close

As discussed in draft-ietf-http-connection-00.txt section 8, in order for an HTTP server to reliably implement the protocol, it needs to shut down 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.

When this feature was added to Apache, it caused a flurry of problems on various versions of Unix because of 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 TCP/IP patches supplied by the vendor. In cases where the vendor has never released patches (i.e., SunOS4 -- although folks with a source license can patch it themselves), we have decided to disable this feature.

There are two ways to accomplish this. One is the socket option SO_LINGER. But as fate would have it, this has never been implemented properly in most TCP/IP stacks. Even 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.

For the most part, Apache implements this in a function called lingering_close (in http_main.c). The function looks roughly like this:

This naturally adds some expense at the end of a connection, but it is required for a reliable implementation. As HTTP/1.1 becomes more 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 NO_LINGCLOSE, but this is not recommended at all. In particular, as HTTP/1.1 pipelined persistent connections come into use, lingering_close is an absolute necessity (and pipelined connections are faster, so you want to support them).

Scoreboard File

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 (and less featured). Peruse the src/main/conf.h file for your architecture, and look for either USE_MMAP_SCOREBOARD or USE_SHMGET_SCOREBOARD. Defining one of those two (as well as their companions HAVE_MMAP and HAVE_SHMGET respectively) enables the supplied shared memory code. If your system has another type of shared memory, edit the file src/main/http_main.c and add the hooks necessary to use it in Apache. (Send us back a patch too, please.)

DYNAMIC_MODULE_LIMIT

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 -DDYNAMIC_MODULE_LIMIT=0 when building your server. This will save RAM that's allocated only for supporting dynamically loaded modules.

Appendix: Detailed Analysis of a Trace

Here is a system call trace of Apache 2.0.38 with the worker MPM on Solaris 8. This trace was collected using:

The -l option tells truss to log the ID of the LWP (lightweight process--Solaris's form of kernel-level thread) that invokes each system call.

Other systems may have different system call tracing utilities such as strace, ktrace, or par. They all produce similar output.

In this trace, a client has requested a 10KB static file from the httpd. Traces of non-static requests or requests with content negotiation look wildly different (and quite ugly in some cases).

/67:    accept(3, 0x00200BEC, 0x00200C0C, 1) (sleeping...)
/67:    accept(3, 0x00200BEC, 0x00200C0C, 1)            = 9

In this trace, the listener thread is running within LWP #67.

/65:    lwp_park(0x00000000, 0)                         = 0
/67:    lwp_unpark(65, 1)                               = 0

Upon accepting the connection, the listener thread wakes up a worker thread to do the request processing. In this trace, the worker thread that handles the request is mapped to LWP #65.

/65:    getsockname(9, 0x00200BA4, 0x00200BC4, 1)       = 0

In order to implement virtual hosts, Apache needs to know the local socket address used to accept the connection. It is possible to eliminate this call in many situations (such as when there are no virtual hosts, or when Listen directives are used which do not have wildcard addresses). But no effort has yet been made to do these optimizations.

/65:    brk(0x002170E8)                                 = 0
/65:    brk(0x002190E8)                                 = 0

The brk(2) calls allocate memory from the heap. It is rare to see these in a system call trace, because the httpd uses custom memory allocators (apr_pool and apr_bucket_alloc) for most request processing. In this trace, the httpd has just been started, so it must call malloc(3) to get the blocks of raw memory with which to create the custom memory allocators.

/65:    fcntl(9, F_GETFL, 0x00000000)                   = 2
/65:    fstat64(9, 0xFAF7B818)                          = 0
/65:    getsockopt(9, 65535, 8192, 0xFAF7B918, 0xFAF7B910, 2190656) = 0
/65:    fstat64(9, 0xFAF7B818)                          = 0
/65:    getsockopt(9, 65535, 8192, 0xFAF7B918, 0xFAF7B914, 2190656) = 0
/65:    setsockopt(9, 65535, 8192, 0xFAF7B918, 4, 2190656) = 0
/65:    fcntl(9, F_SETFL, 0x00000082)                   = 0

Next, the worker thread puts the connection to the client (file descriptor 9) in non-blocking mode. The setsockopt(2) and getsockopt(2) calls are a side-effect of how Solaris's libc handles fcntl(2) on sockets.

/65:    read(9, " G E T   / 1 0 k . h t m".., 8000)     = 97

The worker thread reads the request from the client.

/65:    stat("/var/httpd/apache/httpd-8999/htdocs/10k.html", 0xFAF7B978) = 0
/65:    open("/var/httpd/apache/httpd-8999/htdocs/10k.html", O_RDONLY) = 10

This httpd has been configured with Options FollowSymLinks and AllowOverride None. Thus it doesn't need to lstat(2) each directory in the path leading up to the requested file, nor check for .htaccess files. It simply calls stat(2) to verify that the file: 1) exists, and 2) is a regular file, not a directory.

/65:    sendfilev(0, 9, 0x00200F90, 2, 0xFAF7B53C)      = 10269

In this example, the httpd is able to send the HTTP response header and the requested file with a single sendfilev(2) system call. Sendfile semantics vary among operating systems. On some other systems, it is necessary to do a write(2) or writev(2) call to send the headers before calling sendfile(2).

/65:    write(4, " 1 2 7 . 0 . 0 . 1   -  ".., 78)      = 78

This write(2) call records the request in the access log. Note that one thing missing from this trace is a time(2) call. Unlike Apache 1.3, Apache 2.x uses gettimeofday(3) to look up the time. On some operating systems, like Linux or Solaris, gettimeofday has an optimized implementation that doesn't require as much overhead as a typical system call.

/65:    shutdown(9, 1, 1)                               = 0
/65:    poll(0xFAF7B980, 1, 2000)                       = 1
/65:    read(9, 0xFAF7BC20, 512)                        = 0
/65:    close(9)                                        = 0

The worker thread does a lingering close of the connection.

/65:    close(10)                                       = 0
/65:    lwp_park(0x00000000, 0)         (sleeping...)

Finally the worker thread closes the file that it has just delivered and blocks until the listener assigns it another connection.

/67:    accept(3, 0x001FEB74, 0x001FEB94, 1) (sleeping...)

Meanwhile, the listener thread is able to accept another connection as soon as it has dispatched this connection to a worker thread (subject to some flow-control logic in the worker MPM that throttles the listener if all the available workers are busy). Though it isn't apparent from this trace, the next accept(2) can (and usually does, under high load conditions) occur in parallel with the worker thread's handling of the just-accepted connection.